Upload DixtralForConditionalGeneration
Browse files- README.md +199 -0
- config.json +92 -0
- configuration_dixtral.py +67 -0
- generation_config.json +6 -0
- model-00001-of-00002.safetensors +3 -0
- model-00002-of-00002.safetensors +3 -0
- model.safetensors.index.json +1034 -0
- modeling_dixtral.py +923 -0
README.md
ADDED
|
@@ -0,0 +1,199 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
tags: []
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
# Model Card for Model ID
|
| 7 |
+
|
| 8 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
## Model Details
|
| 13 |
+
|
| 14 |
+
### Model Description
|
| 15 |
+
|
| 16 |
+
<!-- Provide a longer summary of what this model is. -->
|
| 17 |
+
|
| 18 |
+
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
|
| 19 |
+
|
| 20 |
+
- **Developed by:** [More Information Needed]
|
| 21 |
+
- **Funded by [optional]:** [More Information Needed]
|
| 22 |
+
- **Shared by [optional]:** [More Information Needed]
|
| 23 |
+
- **Model type:** [More Information Needed]
|
| 24 |
+
- **Language(s) (NLP):** [More Information Needed]
|
| 25 |
+
- **License:** [More Information Needed]
|
| 26 |
+
- **Finetuned from model [optional]:** [More Information Needed]
|
| 27 |
+
|
| 28 |
+
### Model Sources [optional]
|
| 29 |
+
|
| 30 |
+
<!-- Provide the basic links for the model. -->
|
| 31 |
+
|
| 32 |
+
- **Repository:** [More Information Needed]
|
| 33 |
+
- **Paper [optional]:** [More Information Needed]
|
| 34 |
+
- **Demo [optional]:** [More Information Needed]
|
| 35 |
+
|
| 36 |
+
## Uses
|
| 37 |
+
|
| 38 |
+
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
| 39 |
+
|
| 40 |
+
### Direct Use
|
| 41 |
+
|
| 42 |
+
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
| 43 |
+
|
| 44 |
+
[More Information Needed]
|
| 45 |
+
|
| 46 |
+
### Downstream Use [optional]
|
| 47 |
+
|
| 48 |
+
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
|
| 49 |
+
|
| 50 |
+
[More Information Needed]
|
| 51 |
+
|
| 52 |
+
### Out-of-Scope Use
|
| 53 |
+
|
| 54 |
+
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
|
| 55 |
+
|
| 56 |
+
[More Information Needed]
|
| 57 |
+
|
| 58 |
+
## Bias, Risks, and Limitations
|
| 59 |
+
|
| 60 |
+
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
|
| 61 |
+
|
| 62 |
+
[More Information Needed]
|
| 63 |
+
|
| 64 |
+
### Recommendations
|
| 65 |
+
|
| 66 |
+
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
|
| 67 |
+
|
| 68 |
+
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
|
| 69 |
+
|
| 70 |
+
## How to Get Started with the Model
|
| 71 |
+
|
| 72 |
+
Use the code below to get started with the model.
|
| 73 |
+
|
| 74 |
+
[More Information Needed]
|
| 75 |
+
|
| 76 |
+
## Training Details
|
| 77 |
+
|
| 78 |
+
### Training Data
|
| 79 |
+
|
| 80 |
+
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
|
| 81 |
+
|
| 82 |
+
[More Information Needed]
|
| 83 |
+
|
| 84 |
+
### Training Procedure
|
| 85 |
+
|
| 86 |
+
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
| 87 |
+
|
| 88 |
+
#### Preprocessing [optional]
|
| 89 |
+
|
| 90 |
+
[More Information Needed]
|
| 91 |
+
|
| 92 |
+
|
| 93 |
+
#### Training Hyperparameters
|
| 94 |
+
|
| 95 |
+
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
|
| 96 |
+
|
| 97 |
+
#### Speeds, Sizes, Times [optional]
|
| 98 |
+
|
| 99 |
+
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
| 100 |
+
|
| 101 |
+
[More Information Needed]
|
| 102 |
+
|
| 103 |
+
## Evaluation
|
| 104 |
+
|
| 105 |
+
<!-- This section describes the evaluation protocols and provides the results. -->
|
| 106 |
+
|
| 107 |
+
### Testing Data, Factors & Metrics
|
| 108 |
+
|
| 109 |
+
#### Testing Data
|
| 110 |
+
|
| 111 |
+
<!-- This should link to a Dataset Card if possible. -->
|
| 112 |
+
|
| 113 |
+
[More Information Needed]
|
| 114 |
+
|
| 115 |
+
#### Factors
|
| 116 |
+
|
| 117 |
+
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
| 118 |
+
|
| 119 |
+
[More Information Needed]
|
| 120 |
+
|
| 121 |
+
#### Metrics
|
| 122 |
+
|
| 123 |
+
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
| 124 |
+
|
| 125 |
+
[More Information Needed]
|
| 126 |
+
|
| 127 |
+
### Results
|
| 128 |
+
|
| 129 |
+
[More Information Needed]
|
| 130 |
+
|
| 131 |
+
#### Summary
|
| 132 |
+
|
| 133 |
+
|
| 134 |
+
|
| 135 |
+
## Model Examination [optional]
|
| 136 |
+
|
| 137 |
+
<!-- Relevant interpretability work for the model goes here -->
|
| 138 |
+
|
| 139 |
+
[More Information Needed]
|
| 140 |
+
|
| 141 |
+
## Environmental Impact
|
| 142 |
+
|
| 143 |
+
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
|
| 144 |
+
|
| 145 |
+
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
| 146 |
+
|
| 147 |
+
- **Hardware Type:** [More Information Needed]
|
| 148 |
+
- **Hours used:** [More Information Needed]
|
| 149 |
+
- **Cloud Provider:** [More Information Needed]
|
| 150 |
+
- **Compute Region:** [More Information Needed]
|
| 151 |
+
- **Carbon Emitted:** [More Information Needed]
|
| 152 |
+
|
| 153 |
+
## Technical Specifications [optional]
|
| 154 |
+
|
| 155 |
+
### Model Architecture and Objective
|
| 156 |
+
|
| 157 |
+
[More Information Needed]
|
| 158 |
+
|
| 159 |
+
### Compute Infrastructure
|
| 160 |
+
|
| 161 |
+
[More Information Needed]
|
| 162 |
+
|
| 163 |
+
#### Hardware
|
| 164 |
+
|
| 165 |
+
[More Information Needed]
|
| 166 |
+
|
| 167 |
+
#### Software
|
| 168 |
+
|
| 169 |
+
[More Information Needed]
|
| 170 |
+
|
| 171 |
+
## Citation [optional]
|
| 172 |
+
|
| 173 |
+
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
|
| 174 |
+
|
| 175 |
+
**BibTeX:**
|
| 176 |
+
|
| 177 |
+
[More Information Needed]
|
| 178 |
+
|
| 179 |
+
**APA:**
|
| 180 |
+
|
| 181 |
+
[More Information Needed]
|
| 182 |
+
|
| 183 |
+
## Glossary [optional]
|
| 184 |
+
|
| 185 |
+
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
|
| 186 |
+
|
| 187 |
+
[More Information Needed]
|
| 188 |
+
|
| 189 |
+
## More Information [optional]
|
| 190 |
+
|
| 191 |
+
[More Information Needed]
|
| 192 |
+
|
| 193 |
+
## Model Card Authors [optional]
|
| 194 |
+
|
| 195 |
+
[More Information Needed]
|
| 196 |
+
|
| 197 |
+
## Model Card Contact
|
| 198 |
+
|
| 199 |
+
[More Information Needed]
|
config.json
ADDED
|
@@ -0,0 +1,92 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"DixtralForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"audio_config": {
|
| 6 |
+
"activation_dropout": 0.0,
|
| 7 |
+
"activation_function": "gelu",
|
| 8 |
+
"additional_layer": false,
|
| 9 |
+
"additional_self_attention_layer": false,
|
| 10 |
+
"apply_fddt_to_n_layers": -1,
|
| 11 |
+
"architectures": [
|
| 12 |
+
"DiCoWForConditionalGeneration"
|
| 13 |
+
],
|
| 14 |
+
"attention_dropout": 0.0,
|
| 15 |
+
"bos_token_id": 50257,
|
| 16 |
+
"ctc_loss_reduction": "mean",
|
| 17 |
+
"ctc_weight": 0.0,
|
| 18 |
+
"decoder_start_token_id": 50258,
|
| 19 |
+
"dropout": 0.0,
|
| 20 |
+
"eos_token_id": 50257,
|
| 21 |
+
"fddt_bias_only": false,
|
| 22 |
+
"fddt_init": "suppressive",
|
| 23 |
+
"fddt_is_diagonal": true,
|
| 24 |
+
"fddt_use_non_target": true,
|
| 25 |
+
"fddt_use_overlap": true,
|
| 26 |
+
"fddt_use_silence": true,
|
| 27 |
+
"fddt_use_target": true,
|
| 28 |
+
"final_dropout": 0.0,
|
| 29 |
+
"head_dim": 64,
|
| 30 |
+
"hidden_size": 1280,
|
| 31 |
+
"initializer_range": 0.02,
|
| 32 |
+
"intermediate_size": 5120,
|
| 33 |
+
"is_encoder_decoder": true,
|
| 34 |
+
"layerdrop": 0.0,
|
| 35 |
+
"max_length": null,
|
| 36 |
+
"max_source_positions": 1500,
|
| 37 |
+
"model_type": "voxtral_encoder",
|
| 38 |
+
"non_target_fddt_value": 0.5,
|
| 39 |
+
"num_attention_heads": 20,
|
| 40 |
+
"num_hidden_layers": 32,
|
| 41 |
+
"num_key_value_heads": 20,
|
| 42 |
+
"num_mel_bins": 128,
|
| 43 |
+
"pad_token_id": 50256,
|
| 44 |
+
"pre_ctc_sub_sample": false,
|
| 45 |
+
"remove_timestamps_from_ctc": true,
|
| 46 |
+
"scale_embedding": false,
|
| 47 |
+
"scb_layers": null,
|
| 48 |
+
"torch_dtype": "float32",
|
| 49 |
+
"use_dicow_encoder": true,
|
| 50 |
+
"use_enrollments": false,
|
| 51 |
+
"use_fddt": true,
|
| 52 |
+
"use_pre_pos_fddt": true,
|
| 53 |
+
"vocab_size": 51866
|
| 54 |
+
},
|
| 55 |
+
"audio_token_id": 24,
|
| 56 |
+
"auto_map": {
|
| 57 |
+
"AutoConfig": "configuration_dixtral.DixtralConfig",
|
| 58 |
+
"AutoModel": "modeling_dixtral.DixtralForConditionalGeneration"
|
| 59 |
+
},
|
| 60 |
+
"forced_decoder_ids": null,
|
| 61 |
+
"hidden_size": 3072,
|
| 62 |
+
"model_type": "voxtral",
|
| 63 |
+
"num_soft_prompts": 0,
|
| 64 |
+
"projector_hidden_act": "gelu",
|
| 65 |
+
"text_config": {
|
| 66 |
+
"attention_bias": false,
|
| 67 |
+
"attention_dropout": 0.0,
|
| 68 |
+
"head_dim": 128,
|
| 69 |
+
"hidden_act": "silu",
|
| 70 |
+
"hidden_size": 3072,
|
| 71 |
+
"initializer_range": 0.02,
|
| 72 |
+
"intermediate_size": 8192,
|
| 73 |
+
"max_position_embeddings": 131072,
|
| 74 |
+
"mlp_bias": false,
|
| 75 |
+
"model_type": "llama",
|
| 76 |
+
"num_attention_heads": 32,
|
| 77 |
+
"num_hidden_layers": 30,
|
| 78 |
+
"num_key_value_heads": 8,
|
| 79 |
+
"pretraining_tp": 1,
|
| 80 |
+
"rms_norm_eps": 1e-05,
|
| 81 |
+
"rope_scaling": null,
|
| 82 |
+
"rope_theta": 100000000.0,
|
| 83 |
+
"sliding_window": null,
|
| 84 |
+
"torch_dtype": "float32",
|
| 85 |
+
"use_cache": true,
|
| 86 |
+
"vocab_size": 131072
|
| 87 |
+
},
|
| 88 |
+
"tie_word_embeddings": false,
|
| 89 |
+
"torch_dtype": "bfloat16",
|
| 90 |
+
"transformers_version": "4.55.0",
|
| 91 |
+
"vocab_size": 131072
|
| 92 |
+
}
|
configuration_dixtral.py
ADDED
|
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from transformers.models.voxtral.configuration_voxtral import VoxtralConfig, VoxtralEncoderConfig
|
| 2 |
+
|
| 3 |
+
|
| 4 |
+
class DixtralEncoderConfig(VoxtralEncoderConfig):
|
| 5 |
+
def __init__(
|
| 6 |
+
self,
|
| 7 |
+
# DiCoW-specific parameters
|
| 8 |
+
use_dicow_encoder: bool = False,
|
| 9 |
+
ctc_weight: float = 0.0,
|
| 10 |
+
additional_layer: bool = False,
|
| 11 |
+
additional_self_attention_layer: bool = False,
|
| 12 |
+
pre_ctc_sub_sample: bool = False,
|
| 13 |
+
final_dropout: float = 0.0,
|
| 14 |
+
use_fddt: bool = False,
|
| 15 |
+
apply_fddt_to_n_layers: int = -1,
|
| 16 |
+
use_pre_pos_fddt: bool = False,
|
| 17 |
+
fddt_init: str = "zeros",
|
| 18 |
+
fddt_is_diagonal: bool = False,
|
| 19 |
+
fddt_bias_only: bool = False,
|
| 20 |
+
fddt_use_silence: bool = True,
|
| 21 |
+
fddt_use_target: bool = True,
|
| 22 |
+
fddt_use_overlap: bool = True,
|
| 23 |
+
fddt_use_non_target: bool = True,
|
| 24 |
+
non_target_fddt_value: float = 1.0,
|
| 25 |
+
use_enrollments: bool = False,
|
| 26 |
+
scb_layers: int = None,
|
| 27 |
+
remove_timestamps_from_ctc: bool = False,
|
| 28 |
+
ctc_loss_reduction: str = "mean",
|
| 29 |
+
**kwargs
|
| 30 |
+
):
|
| 31 |
+
super().__init__(**kwargs)
|
| 32 |
+
self.use_dicow_encoder = use_dicow_encoder
|
| 33 |
+
self.ctc_weight = ctc_weight
|
| 34 |
+
self.additional_layer = additional_layer
|
| 35 |
+
self.additional_self_attention_layer = additional_self_attention_layer
|
| 36 |
+
self.pre_ctc_sub_sample = pre_ctc_sub_sample
|
| 37 |
+
self.final_dropout = final_dropout
|
| 38 |
+
self.use_fddt = use_fddt
|
| 39 |
+
self.apply_fddt_to_n_layers = apply_fddt_to_n_layers
|
| 40 |
+
self.use_pre_pos_fddt = use_pre_pos_fddt
|
| 41 |
+
self.fddt_init = fddt_init
|
| 42 |
+
self.fddt_is_diagonal = fddt_is_diagonal
|
| 43 |
+
self.fddt_bias_only = fddt_bias_only
|
| 44 |
+
self.fddt_use_silence = fddt_use_silence
|
| 45 |
+
self.fddt_use_target = fddt_use_target
|
| 46 |
+
self.fddt_use_overlap = fddt_use_overlap
|
| 47 |
+
self.fddt_use_non_target = fddt_use_non_target
|
| 48 |
+
self.non_target_fddt_value = non_target_fddt_value
|
| 49 |
+
self.use_enrollments = use_enrollments
|
| 50 |
+
self.scb_layers = scb_layers
|
| 51 |
+
self.remove_timestamps_from_ctc = remove_timestamps_from_ctc
|
| 52 |
+
self.ctc_loss_reduction = ctc_loss_reduction
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
class DixtralConfig(VoxtralConfig):
|
| 56 |
+
def __init__(
|
| 57 |
+
self,
|
| 58 |
+
audio_config: dict = None,
|
| 59 |
+
num_soft_prompts: int = 0,
|
| 60 |
+
**kwargs
|
| 61 |
+
):
|
| 62 |
+
# Convert audio_config to DiCoW version if provided
|
| 63 |
+
if audio_config is not None and not isinstance(audio_config, DixtralEncoderConfig):
|
| 64 |
+
audio_config = DixtralEncoderConfig(**audio_config)
|
| 65 |
+
|
| 66 |
+
super().__init__(audio_config=audio_config, **kwargs)
|
| 67 |
+
self.num_soft_prompts = num_soft_prompts
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 1,
|
| 3 |
+
"eos_token_id": 2,
|
| 4 |
+
"pad_token_id": 11,
|
| 5 |
+
"transformers_version": "4.55.0"
|
| 6 |
+
}
|
model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ce5ab478731b655f4b01ddc686ec185343098394f95ef9e9ecbb77f6de01cfb5
|
| 3 |
+
size 4974251192
|
model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e57cd9268d98f4fb0d2950c7fbe8c42b78498f57a246f18f131f58849c06667d
|
| 3 |
+
size 4379088336
|
model.safetensors.index.json
ADDED
|
@@ -0,0 +1,1034 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"metadata": {
|
| 3 |
+
"total_parameters": 4676609024,
|
| 4 |
+
"total_size": 9353218048
|
| 5 |
+
},
|
| 6 |
+
"weight_map": {
|
| 7 |
+
"audio_tower.conv1.bias": "model-00001-of-00002.safetensors",
|
| 8 |
+
"audio_tower.conv1.weight": "model-00001-of-00002.safetensors",
|
| 9 |
+
"audio_tower.conv2.bias": "model-00001-of-00002.safetensors",
|
| 10 |
+
"audio_tower.conv2.weight": "model-00001-of-00002.safetensors",
|
| 11 |
+
"audio_tower.embed_positions.weight": "model-00001-of-00002.safetensors",
|
| 12 |
+
"audio_tower.fddts.0.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 13 |
+
"audio_tower.fddts.0.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 14 |
+
"audio_tower.fddts.0.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 15 |
+
"audio_tower.fddts.0.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 16 |
+
"audio_tower.fddts.0.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 17 |
+
"audio_tower.fddts.0.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 18 |
+
"audio_tower.fddts.0.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 19 |
+
"audio_tower.fddts.0.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 20 |
+
"audio_tower.fddts.1.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 21 |
+
"audio_tower.fddts.1.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 22 |
+
"audio_tower.fddts.1.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 23 |
+
"audio_tower.fddts.1.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 24 |
+
"audio_tower.fddts.1.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 25 |
+
"audio_tower.fddts.1.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 26 |
+
"audio_tower.fddts.1.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 27 |
+
"audio_tower.fddts.1.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 28 |
+
"audio_tower.fddts.10.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 29 |
+
"audio_tower.fddts.10.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 30 |
+
"audio_tower.fddts.10.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 31 |
+
"audio_tower.fddts.10.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 32 |
+
"audio_tower.fddts.10.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 33 |
+
"audio_tower.fddts.10.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 34 |
+
"audio_tower.fddts.10.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 35 |
+
"audio_tower.fddts.10.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 36 |
+
"audio_tower.fddts.11.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 37 |
+
"audio_tower.fddts.11.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 38 |
+
"audio_tower.fddts.11.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 39 |
+
"audio_tower.fddts.11.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 40 |
+
"audio_tower.fddts.11.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 41 |
+
"audio_tower.fddts.11.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 42 |
+
"audio_tower.fddts.11.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 43 |
+
"audio_tower.fddts.11.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 44 |
+
"audio_tower.fddts.12.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 45 |
+
"audio_tower.fddts.12.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 46 |
+
"audio_tower.fddts.12.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 47 |
+
"audio_tower.fddts.12.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 48 |
+
"audio_tower.fddts.12.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 49 |
+
"audio_tower.fddts.12.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 50 |
+
"audio_tower.fddts.12.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 51 |
+
"audio_tower.fddts.12.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 52 |
+
"audio_tower.fddts.13.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 53 |
+
"audio_tower.fddts.13.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 54 |
+
"audio_tower.fddts.13.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 55 |
+
"audio_tower.fddts.13.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 56 |
+
"audio_tower.fddts.13.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 57 |
+
"audio_tower.fddts.13.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 58 |
+
"audio_tower.fddts.13.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 59 |
+
"audio_tower.fddts.13.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 60 |
+
"audio_tower.fddts.14.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 61 |
+
"audio_tower.fddts.14.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 62 |
+
"audio_tower.fddts.14.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 63 |
+
"audio_tower.fddts.14.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 64 |
+
"audio_tower.fddts.14.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 65 |
+
"audio_tower.fddts.14.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 66 |
+
"audio_tower.fddts.14.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 67 |
+
"audio_tower.fddts.14.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 68 |
+
"audio_tower.fddts.15.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 69 |
+
"audio_tower.fddts.15.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 70 |
+
"audio_tower.fddts.15.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 71 |
+
"audio_tower.fddts.15.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 72 |
+
"audio_tower.fddts.15.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 73 |
+
"audio_tower.fddts.15.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 74 |
+
"audio_tower.fddts.15.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 75 |
+
"audio_tower.fddts.15.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 76 |
+
"audio_tower.fddts.16.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 77 |
+
"audio_tower.fddts.16.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 78 |
+
"audio_tower.fddts.16.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 79 |
+
"audio_tower.fddts.16.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 80 |
+
"audio_tower.fddts.16.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 81 |
+
"audio_tower.fddts.16.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 82 |
+
"audio_tower.fddts.16.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 83 |
+
"audio_tower.fddts.16.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 84 |
+
"audio_tower.fddts.17.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 85 |
+
"audio_tower.fddts.17.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 86 |
+
"audio_tower.fddts.17.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 87 |
+
"audio_tower.fddts.17.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 88 |
+
"audio_tower.fddts.17.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 89 |
+
"audio_tower.fddts.17.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 90 |
+
"audio_tower.fddts.17.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 91 |
+
"audio_tower.fddts.17.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 92 |
+
"audio_tower.fddts.18.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 93 |
+
"audio_tower.fddts.18.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 94 |
+
"audio_tower.fddts.18.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 95 |
+
"audio_tower.fddts.18.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 96 |
+
"audio_tower.fddts.18.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 97 |
+
"audio_tower.fddts.18.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 98 |
+
"audio_tower.fddts.18.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 99 |
+
"audio_tower.fddts.18.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 100 |
+
"audio_tower.fddts.19.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 101 |
+
"audio_tower.fddts.19.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 102 |
+
"audio_tower.fddts.19.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 103 |
+
"audio_tower.fddts.19.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 104 |
+
"audio_tower.fddts.19.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 105 |
+
"audio_tower.fddts.19.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 106 |
+
"audio_tower.fddts.19.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 107 |
+
"audio_tower.fddts.19.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 108 |
+
"audio_tower.fddts.2.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 109 |
+
"audio_tower.fddts.2.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 110 |
+
"audio_tower.fddts.2.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 111 |
+
"audio_tower.fddts.2.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 112 |
+
"audio_tower.fddts.2.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 113 |
+
"audio_tower.fddts.2.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 114 |
+
"audio_tower.fddts.2.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 115 |
+
"audio_tower.fddts.2.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 116 |
+
"audio_tower.fddts.20.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 117 |
+
"audio_tower.fddts.20.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 118 |
+
"audio_tower.fddts.20.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 119 |
+
"audio_tower.fddts.20.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 120 |
+
"audio_tower.fddts.20.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 121 |
+
"audio_tower.fddts.20.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 122 |
+
"audio_tower.fddts.20.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 123 |
+
"audio_tower.fddts.20.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 124 |
+
"audio_tower.fddts.21.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 125 |
+
"audio_tower.fddts.21.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 126 |
+
"audio_tower.fddts.21.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 127 |
+
"audio_tower.fddts.21.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 128 |
+
"audio_tower.fddts.21.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 129 |
+
"audio_tower.fddts.21.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 130 |
+
"audio_tower.fddts.21.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 131 |
+
"audio_tower.fddts.21.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 132 |
+
"audio_tower.fddts.22.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 133 |
+
"audio_tower.fddts.22.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 134 |
+
"audio_tower.fddts.22.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 135 |
+
"audio_tower.fddts.22.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 136 |
+
"audio_tower.fddts.22.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 137 |
+
"audio_tower.fddts.22.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 138 |
+
"audio_tower.fddts.22.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 139 |
+
"audio_tower.fddts.22.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 140 |
+
"audio_tower.fddts.23.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 141 |
+
"audio_tower.fddts.23.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 142 |
+
"audio_tower.fddts.23.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 143 |
+
"audio_tower.fddts.23.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 144 |
+
"audio_tower.fddts.23.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 145 |
+
"audio_tower.fddts.23.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 146 |
+
"audio_tower.fddts.23.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 147 |
+
"audio_tower.fddts.23.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 148 |
+
"audio_tower.fddts.24.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 149 |
+
"audio_tower.fddts.24.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 150 |
+
"audio_tower.fddts.24.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 151 |
+
"audio_tower.fddts.24.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 152 |
+
"audio_tower.fddts.24.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 153 |
+
"audio_tower.fddts.24.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 154 |
+
"audio_tower.fddts.24.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 155 |
+
"audio_tower.fddts.24.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 156 |
+
"audio_tower.fddts.25.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 157 |
+
"audio_tower.fddts.25.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 158 |
+
"audio_tower.fddts.25.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 159 |
+
"audio_tower.fddts.25.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 160 |
+
"audio_tower.fddts.25.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 161 |
+
"audio_tower.fddts.25.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 162 |
+
"audio_tower.fddts.25.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 163 |
+
"audio_tower.fddts.25.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 164 |
+
"audio_tower.fddts.26.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 165 |
+
"audio_tower.fddts.26.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 166 |
+
"audio_tower.fddts.26.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 167 |
+
"audio_tower.fddts.26.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 168 |
+
"audio_tower.fddts.26.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 169 |
+
"audio_tower.fddts.26.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 170 |
+
"audio_tower.fddts.26.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 171 |
+
"audio_tower.fddts.26.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 172 |
+
"audio_tower.fddts.27.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 173 |
+
"audio_tower.fddts.27.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 174 |
+
"audio_tower.fddts.27.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 175 |
+
"audio_tower.fddts.27.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 176 |
+
"audio_tower.fddts.27.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 177 |
+
"audio_tower.fddts.27.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 178 |
+
"audio_tower.fddts.27.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 179 |
+
"audio_tower.fddts.27.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 180 |
+
"audio_tower.fddts.28.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 181 |
+
"audio_tower.fddts.28.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 182 |
+
"audio_tower.fddts.28.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 183 |
+
"audio_tower.fddts.28.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 184 |
+
"audio_tower.fddts.28.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 185 |
+
"audio_tower.fddts.28.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 186 |
+
"audio_tower.fddts.28.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 187 |
+
"audio_tower.fddts.28.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 188 |
+
"audio_tower.fddts.29.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 189 |
+
"audio_tower.fddts.29.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 190 |
+
"audio_tower.fddts.29.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 191 |
+
"audio_tower.fddts.29.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 192 |
+
"audio_tower.fddts.29.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 193 |
+
"audio_tower.fddts.29.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 194 |
+
"audio_tower.fddts.29.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 195 |
+
"audio_tower.fddts.29.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 196 |
+
"audio_tower.fddts.3.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 197 |
+
"audio_tower.fddts.3.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 198 |
+
"audio_tower.fddts.3.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 199 |
+
"audio_tower.fddts.3.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 200 |
+
"audio_tower.fddts.3.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 201 |
+
"audio_tower.fddts.3.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 202 |
+
"audio_tower.fddts.3.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 203 |
+
"audio_tower.fddts.3.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 204 |
+
"audio_tower.fddts.30.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 205 |
+
"audio_tower.fddts.30.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 206 |
+
"audio_tower.fddts.30.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 207 |
+
"audio_tower.fddts.30.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 208 |
+
"audio_tower.fddts.30.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 209 |
+
"audio_tower.fddts.30.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 210 |
+
"audio_tower.fddts.30.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 211 |
+
"audio_tower.fddts.30.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 212 |
+
"audio_tower.fddts.31.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 213 |
+
"audio_tower.fddts.31.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 214 |
+
"audio_tower.fddts.31.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 215 |
+
"audio_tower.fddts.31.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 216 |
+
"audio_tower.fddts.31.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 217 |
+
"audio_tower.fddts.31.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 218 |
+
"audio_tower.fddts.31.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 219 |
+
"audio_tower.fddts.31.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 220 |
+
"audio_tower.fddts.4.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 221 |
+
"audio_tower.fddts.4.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 222 |
+
"audio_tower.fddts.4.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 223 |
+
"audio_tower.fddts.4.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 224 |
+
"audio_tower.fddts.4.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 225 |
+
"audio_tower.fddts.4.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 226 |
+
"audio_tower.fddts.4.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 227 |
+
"audio_tower.fddts.4.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 228 |
+
"audio_tower.fddts.5.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 229 |
+
"audio_tower.fddts.5.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 230 |
+
"audio_tower.fddts.5.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 231 |
+
"audio_tower.fddts.5.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 232 |
+
"audio_tower.fddts.5.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 233 |
+
"audio_tower.fddts.5.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 234 |
+
"audio_tower.fddts.5.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 235 |
+
"audio_tower.fddts.5.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 236 |
+
"audio_tower.fddts.6.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 237 |
+
"audio_tower.fddts.6.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 238 |
+
"audio_tower.fddts.6.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 239 |
+
"audio_tower.fddts.6.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 240 |
+
"audio_tower.fddts.6.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 241 |
+
"audio_tower.fddts.6.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 242 |
+
"audio_tower.fddts.6.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 243 |
+
"audio_tower.fddts.6.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 244 |
+
"audio_tower.fddts.7.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 245 |
+
"audio_tower.fddts.7.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 246 |
+
"audio_tower.fddts.7.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 247 |
+
"audio_tower.fddts.7.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 248 |
+
"audio_tower.fddts.7.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 249 |
+
"audio_tower.fddts.7.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 250 |
+
"audio_tower.fddts.7.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 251 |
+
"audio_tower.fddts.7.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 252 |
+
"audio_tower.fddts.8.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 253 |
+
"audio_tower.fddts.8.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 254 |
+
"audio_tower.fddts.8.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 255 |
+
"audio_tower.fddts.8.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 256 |
+
"audio_tower.fddts.8.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 257 |
+
"audio_tower.fddts.8.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 258 |
+
"audio_tower.fddts.8.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 259 |
+
"audio_tower.fddts.8.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 260 |
+
"audio_tower.fddts.9.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 261 |
+
"audio_tower.fddts.9.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 262 |
+
"audio_tower.fddts.9.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 263 |
+
"audio_tower.fddts.9.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 264 |
+
"audio_tower.fddts.9.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 265 |
+
"audio_tower.fddts.9.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 266 |
+
"audio_tower.fddts.9.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 267 |
+
"audio_tower.fddts.9.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 268 |
+
"audio_tower.initial_fddt.non_target_linear.bias": "model-00001-of-00002.safetensors",
|
| 269 |
+
"audio_tower.initial_fddt.non_target_linear.weight": "model-00001-of-00002.safetensors",
|
| 270 |
+
"audio_tower.initial_fddt.overlap_linear.bias": "model-00001-of-00002.safetensors",
|
| 271 |
+
"audio_tower.initial_fddt.overlap_linear.weight": "model-00001-of-00002.safetensors",
|
| 272 |
+
"audio_tower.initial_fddt.silence_linear.bias": "model-00001-of-00002.safetensors",
|
| 273 |
+
"audio_tower.initial_fddt.silence_linear.weight": "model-00001-of-00002.safetensors",
|
| 274 |
+
"audio_tower.initial_fddt.target_linear.bias": "model-00001-of-00002.safetensors",
|
| 275 |
+
"audio_tower.initial_fddt.target_linear.weight": "model-00001-of-00002.safetensors",
|
| 276 |
+
"audio_tower.layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 277 |
+
"audio_tower.layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 278 |
+
"audio_tower.layers.0.fc1.bias": "model-00001-of-00002.safetensors",
|
| 279 |
+
"audio_tower.layers.0.fc1.weight": "model-00001-of-00002.safetensors",
|
| 280 |
+
"audio_tower.layers.0.fc2.bias": "model-00001-of-00002.safetensors",
|
| 281 |
+
"audio_tower.layers.0.fc2.weight": "model-00001-of-00002.safetensors",
|
| 282 |
+
"audio_tower.layers.0.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 283 |
+
"audio_tower.layers.0.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 284 |
+
"audio_tower.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 285 |
+
"audio_tower.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 286 |
+
"audio_tower.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 287 |
+
"audio_tower.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 288 |
+
"audio_tower.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 289 |
+
"audio_tower.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 290 |
+
"audio_tower.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 291 |
+
"audio_tower.layers.0.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 292 |
+
"audio_tower.layers.0.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 293 |
+
"audio_tower.layers.1.fc1.bias": "model-00001-of-00002.safetensors",
|
| 294 |
+
"audio_tower.layers.1.fc1.weight": "model-00001-of-00002.safetensors",
|
| 295 |
+
"audio_tower.layers.1.fc2.bias": "model-00001-of-00002.safetensors",
|
| 296 |
+
"audio_tower.layers.1.fc2.weight": "model-00001-of-00002.safetensors",
|
| 297 |
+
"audio_tower.layers.1.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 298 |
+
"audio_tower.layers.1.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 299 |
+
"audio_tower.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 300 |
+
"audio_tower.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 301 |
+
"audio_tower.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 302 |
+
"audio_tower.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 303 |
+
"audio_tower.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 304 |
+
"audio_tower.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 305 |
+
"audio_tower.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 306 |
+
"audio_tower.layers.1.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 307 |
+
"audio_tower.layers.1.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 308 |
+
"audio_tower.layers.10.fc1.bias": "model-00001-of-00002.safetensors",
|
| 309 |
+
"audio_tower.layers.10.fc1.weight": "model-00001-of-00002.safetensors",
|
| 310 |
+
"audio_tower.layers.10.fc2.bias": "model-00001-of-00002.safetensors",
|
| 311 |
+
"audio_tower.layers.10.fc2.weight": "model-00001-of-00002.safetensors",
|
| 312 |
+
"audio_tower.layers.10.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 313 |
+
"audio_tower.layers.10.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 314 |
+
"audio_tower.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 315 |
+
"audio_tower.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 316 |
+
"audio_tower.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 317 |
+
"audio_tower.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 318 |
+
"audio_tower.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 319 |
+
"audio_tower.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 320 |
+
"audio_tower.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 321 |
+
"audio_tower.layers.10.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 322 |
+
"audio_tower.layers.10.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 323 |
+
"audio_tower.layers.11.fc1.bias": "model-00001-of-00002.safetensors",
|
| 324 |
+
"audio_tower.layers.11.fc1.weight": "model-00001-of-00002.safetensors",
|
| 325 |
+
"audio_tower.layers.11.fc2.bias": "model-00001-of-00002.safetensors",
|
| 326 |
+
"audio_tower.layers.11.fc2.weight": "model-00001-of-00002.safetensors",
|
| 327 |
+
"audio_tower.layers.11.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 328 |
+
"audio_tower.layers.11.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 329 |
+
"audio_tower.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 330 |
+
"audio_tower.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 331 |
+
"audio_tower.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 332 |
+
"audio_tower.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 333 |
+
"audio_tower.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 334 |
+
"audio_tower.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 335 |
+
"audio_tower.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 336 |
+
"audio_tower.layers.11.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 337 |
+
"audio_tower.layers.11.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 338 |
+
"audio_tower.layers.12.fc1.bias": "model-00001-of-00002.safetensors",
|
| 339 |
+
"audio_tower.layers.12.fc1.weight": "model-00001-of-00002.safetensors",
|
| 340 |
+
"audio_tower.layers.12.fc2.bias": "model-00001-of-00002.safetensors",
|
| 341 |
+
"audio_tower.layers.12.fc2.weight": "model-00001-of-00002.safetensors",
|
| 342 |
+
"audio_tower.layers.12.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 343 |
+
"audio_tower.layers.12.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 344 |
+
"audio_tower.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 345 |
+
"audio_tower.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 346 |
+
"audio_tower.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 347 |
+
"audio_tower.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 348 |
+
"audio_tower.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 349 |
+
"audio_tower.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 350 |
+
"audio_tower.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 351 |
+
"audio_tower.layers.12.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 352 |
+
"audio_tower.layers.12.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 353 |
+
"audio_tower.layers.13.fc1.bias": "model-00001-of-00002.safetensors",
|
| 354 |
+
"audio_tower.layers.13.fc1.weight": "model-00001-of-00002.safetensors",
|
| 355 |
+
"audio_tower.layers.13.fc2.bias": "model-00001-of-00002.safetensors",
|
| 356 |
+
"audio_tower.layers.13.fc2.weight": "model-00001-of-00002.safetensors",
|
| 357 |
+
"audio_tower.layers.13.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 358 |
+
"audio_tower.layers.13.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 359 |
+
"audio_tower.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 360 |
+
"audio_tower.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 361 |
+
"audio_tower.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 362 |
+
"audio_tower.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 363 |
+
"audio_tower.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 364 |
+
"audio_tower.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 365 |
+
"audio_tower.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 366 |
+
"audio_tower.layers.13.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 367 |
+
"audio_tower.layers.13.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 368 |
+
"audio_tower.layers.14.fc1.bias": "model-00001-of-00002.safetensors",
|
| 369 |
+
"audio_tower.layers.14.fc1.weight": "model-00001-of-00002.safetensors",
|
| 370 |
+
"audio_tower.layers.14.fc2.bias": "model-00001-of-00002.safetensors",
|
| 371 |
+
"audio_tower.layers.14.fc2.weight": "model-00001-of-00002.safetensors",
|
| 372 |
+
"audio_tower.layers.14.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 373 |
+
"audio_tower.layers.14.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 374 |
+
"audio_tower.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 375 |
+
"audio_tower.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 376 |
+
"audio_tower.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 377 |
+
"audio_tower.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 378 |
+
"audio_tower.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 379 |
+
"audio_tower.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 380 |
+
"audio_tower.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 381 |
+
"audio_tower.layers.14.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 382 |
+
"audio_tower.layers.14.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 383 |
+
"audio_tower.layers.15.fc1.bias": "model-00001-of-00002.safetensors",
|
| 384 |
+
"audio_tower.layers.15.fc1.weight": "model-00001-of-00002.safetensors",
|
| 385 |
+
"audio_tower.layers.15.fc2.bias": "model-00001-of-00002.safetensors",
|
| 386 |
+
"audio_tower.layers.15.fc2.weight": "model-00001-of-00002.safetensors",
|
| 387 |
+
"audio_tower.layers.15.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 388 |
+
"audio_tower.layers.15.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 389 |
+
"audio_tower.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 390 |
+
"audio_tower.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 391 |
+
"audio_tower.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 392 |
+
"audio_tower.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 393 |
+
"audio_tower.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 394 |
+
"audio_tower.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 395 |
+
"audio_tower.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 396 |
+
"audio_tower.layers.15.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 397 |
+
"audio_tower.layers.15.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 398 |
+
"audio_tower.layers.16.fc1.bias": "model-00001-of-00002.safetensors",
|
| 399 |
+
"audio_tower.layers.16.fc1.weight": "model-00001-of-00002.safetensors",
|
| 400 |
+
"audio_tower.layers.16.fc2.bias": "model-00001-of-00002.safetensors",
|
| 401 |
+
"audio_tower.layers.16.fc2.weight": "model-00001-of-00002.safetensors",
|
| 402 |
+
"audio_tower.layers.16.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 403 |
+
"audio_tower.layers.16.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 404 |
+
"audio_tower.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 405 |
+
"audio_tower.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 406 |
+
"audio_tower.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 407 |
+
"audio_tower.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 408 |
+
"audio_tower.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 409 |
+
"audio_tower.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 410 |
+
"audio_tower.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 411 |
+
"audio_tower.layers.16.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 412 |
+
"audio_tower.layers.16.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 413 |
+
"audio_tower.layers.17.fc1.bias": "model-00001-of-00002.safetensors",
|
| 414 |
+
"audio_tower.layers.17.fc1.weight": "model-00001-of-00002.safetensors",
|
| 415 |
+
"audio_tower.layers.17.fc2.bias": "model-00001-of-00002.safetensors",
|
| 416 |
+
"audio_tower.layers.17.fc2.weight": "model-00001-of-00002.safetensors",
|
| 417 |
+
"audio_tower.layers.17.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 418 |
+
"audio_tower.layers.17.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 419 |
+
"audio_tower.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 420 |
+
"audio_tower.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 421 |
+
"audio_tower.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 422 |
+
"audio_tower.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 423 |
+
"audio_tower.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 424 |
+
"audio_tower.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 425 |
+
"audio_tower.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 426 |
+
"audio_tower.layers.17.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 427 |
+
"audio_tower.layers.17.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 428 |
+
"audio_tower.layers.18.fc1.bias": "model-00001-of-00002.safetensors",
|
| 429 |
+
"audio_tower.layers.18.fc1.weight": "model-00001-of-00002.safetensors",
|
| 430 |
+
"audio_tower.layers.18.fc2.bias": "model-00001-of-00002.safetensors",
|
| 431 |
+
"audio_tower.layers.18.fc2.weight": "model-00001-of-00002.safetensors",
|
| 432 |
+
"audio_tower.layers.18.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 433 |
+
"audio_tower.layers.18.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 434 |
+
"audio_tower.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 435 |
+
"audio_tower.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 436 |
+
"audio_tower.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 437 |
+
"audio_tower.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 438 |
+
"audio_tower.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 439 |
+
"audio_tower.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 440 |
+
"audio_tower.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 441 |
+
"audio_tower.layers.18.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 442 |
+
"audio_tower.layers.18.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 443 |
+
"audio_tower.layers.19.fc1.bias": "model-00001-of-00002.safetensors",
|
| 444 |
+
"audio_tower.layers.19.fc1.weight": "model-00001-of-00002.safetensors",
|
| 445 |
+
"audio_tower.layers.19.fc2.bias": "model-00001-of-00002.safetensors",
|
| 446 |
+
"audio_tower.layers.19.fc2.weight": "model-00001-of-00002.safetensors",
|
| 447 |
+
"audio_tower.layers.19.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 448 |
+
"audio_tower.layers.19.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 449 |
+
"audio_tower.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 450 |
+
"audio_tower.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 451 |
+
"audio_tower.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 452 |
+
"audio_tower.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 453 |
+
"audio_tower.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 454 |
+
"audio_tower.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 455 |
+
"audio_tower.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 456 |
+
"audio_tower.layers.19.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 457 |
+
"audio_tower.layers.19.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 458 |
+
"audio_tower.layers.2.fc1.bias": "model-00001-of-00002.safetensors",
|
| 459 |
+
"audio_tower.layers.2.fc1.weight": "model-00001-of-00002.safetensors",
|
| 460 |
+
"audio_tower.layers.2.fc2.bias": "model-00001-of-00002.safetensors",
|
| 461 |
+
"audio_tower.layers.2.fc2.weight": "model-00001-of-00002.safetensors",
|
| 462 |
+
"audio_tower.layers.2.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 463 |
+
"audio_tower.layers.2.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 464 |
+
"audio_tower.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 465 |
+
"audio_tower.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 466 |
+
"audio_tower.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 467 |
+
"audio_tower.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 468 |
+
"audio_tower.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 469 |
+
"audio_tower.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 470 |
+
"audio_tower.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 471 |
+
"audio_tower.layers.2.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 472 |
+
"audio_tower.layers.2.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 473 |
+
"audio_tower.layers.20.fc1.bias": "model-00001-of-00002.safetensors",
|
| 474 |
+
"audio_tower.layers.20.fc1.weight": "model-00001-of-00002.safetensors",
|
| 475 |
+
"audio_tower.layers.20.fc2.bias": "model-00001-of-00002.safetensors",
|
| 476 |
+
"audio_tower.layers.20.fc2.weight": "model-00001-of-00002.safetensors",
|
| 477 |
+
"audio_tower.layers.20.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 478 |
+
"audio_tower.layers.20.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 479 |
+
"audio_tower.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 480 |
+
"audio_tower.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 481 |
+
"audio_tower.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 482 |
+
"audio_tower.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 483 |
+
"audio_tower.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 484 |
+
"audio_tower.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 485 |
+
"audio_tower.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 486 |
+
"audio_tower.layers.20.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 487 |
+
"audio_tower.layers.20.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 488 |
+
"audio_tower.layers.21.fc1.bias": "model-00001-of-00002.safetensors",
|
| 489 |
+
"audio_tower.layers.21.fc1.weight": "model-00001-of-00002.safetensors",
|
| 490 |
+
"audio_tower.layers.21.fc2.bias": "model-00001-of-00002.safetensors",
|
| 491 |
+
"audio_tower.layers.21.fc2.weight": "model-00001-of-00002.safetensors",
|
| 492 |
+
"audio_tower.layers.21.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 493 |
+
"audio_tower.layers.21.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 494 |
+
"audio_tower.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 495 |
+
"audio_tower.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 496 |
+
"audio_tower.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 497 |
+
"audio_tower.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 498 |
+
"audio_tower.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 499 |
+
"audio_tower.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 500 |
+
"audio_tower.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 501 |
+
"audio_tower.layers.21.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 502 |
+
"audio_tower.layers.21.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 503 |
+
"audio_tower.layers.22.fc1.bias": "model-00001-of-00002.safetensors",
|
| 504 |
+
"audio_tower.layers.22.fc1.weight": "model-00001-of-00002.safetensors",
|
| 505 |
+
"audio_tower.layers.22.fc2.bias": "model-00001-of-00002.safetensors",
|
| 506 |
+
"audio_tower.layers.22.fc2.weight": "model-00001-of-00002.safetensors",
|
| 507 |
+
"audio_tower.layers.22.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 508 |
+
"audio_tower.layers.22.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 509 |
+
"audio_tower.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 510 |
+
"audio_tower.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 511 |
+
"audio_tower.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 512 |
+
"audio_tower.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 513 |
+
"audio_tower.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 514 |
+
"audio_tower.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 515 |
+
"audio_tower.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 516 |
+
"audio_tower.layers.22.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 517 |
+
"audio_tower.layers.22.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 518 |
+
"audio_tower.layers.23.fc1.bias": "model-00001-of-00002.safetensors",
|
| 519 |
+
"audio_tower.layers.23.fc1.weight": "model-00001-of-00002.safetensors",
|
| 520 |
+
"audio_tower.layers.23.fc2.bias": "model-00001-of-00002.safetensors",
|
| 521 |
+
"audio_tower.layers.23.fc2.weight": "model-00001-of-00002.safetensors",
|
| 522 |
+
"audio_tower.layers.23.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 523 |
+
"audio_tower.layers.23.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 524 |
+
"audio_tower.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 525 |
+
"audio_tower.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 526 |
+
"audio_tower.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 527 |
+
"audio_tower.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 528 |
+
"audio_tower.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 529 |
+
"audio_tower.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 530 |
+
"audio_tower.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 531 |
+
"audio_tower.layers.23.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 532 |
+
"audio_tower.layers.23.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 533 |
+
"audio_tower.layers.24.fc1.bias": "model-00001-of-00002.safetensors",
|
| 534 |
+
"audio_tower.layers.24.fc1.weight": "model-00001-of-00002.safetensors",
|
| 535 |
+
"audio_tower.layers.24.fc2.bias": "model-00001-of-00002.safetensors",
|
| 536 |
+
"audio_tower.layers.24.fc2.weight": "model-00001-of-00002.safetensors",
|
| 537 |
+
"audio_tower.layers.24.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 538 |
+
"audio_tower.layers.24.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 539 |
+
"audio_tower.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 540 |
+
"audio_tower.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 541 |
+
"audio_tower.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 542 |
+
"audio_tower.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 543 |
+
"audio_tower.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 544 |
+
"audio_tower.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 545 |
+
"audio_tower.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 546 |
+
"audio_tower.layers.24.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 547 |
+
"audio_tower.layers.24.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 548 |
+
"audio_tower.layers.25.fc1.bias": "model-00001-of-00002.safetensors",
|
| 549 |
+
"audio_tower.layers.25.fc1.weight": "model-00001-of-00002.safetensors",
|
| 550 |
+
"audio_tower.layers.25.fc2.bias": "model-00001-of-00002.safetensors",
|
| 551 |
+
"audio_tower.layers.25.fc2.weight": "model-00001-of-00002.safetensors",
|
| 552 |
+
"audio_tower.layers.25.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 553 |
+
"audio_tower.layers.25.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 554 |
+
"audio_tower.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 555 |
+
"audio_tower.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 556 |
+
"audio_tower.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 557 |
+
"audio_tower.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 558 |
+
"audio_tower.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 559 |
+
"audio_tower.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 560 |
+
"audio_tower.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 561 |
+
"audio_tower.layers.25.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 562 |
+
"audio_tower.layers.25.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 563 |
+
"audio_tower.layers.26.fc1.bias": "model-00001-of-00002.safetensors",
|
| 564 |
+
"audio_tower.layers.26.fc1.weight": "model-00001-of-00002.safetensors",
|
| 565 |
+
"audio_tower.layers.26.fc2.bias": "model-00001-of-00002.safetensors",
|
| 566 |
+
"audio_tower.layers.26.fc2.weight": "model-00001-of-00002.safetensors",
|
| 567 |
+
"audio_tower.layers.26.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 568 |
+
"audio_tower.layers.26.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 569 |
+
"audio_tower.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 570 |
+
"audio_tower.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 571 |
+
"audio_tower.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 572 |
+
"audio_tower.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 573 |
+
"audio_tower.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 574 |
+
"audio_tower.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 575 |
+
"audio_tower.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 576 |
+
"audio_tower.layers.26.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 577 |
+
"audio_tower.layers.26.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 578 |
+
"audio_tower.layers.27.fc1.bias": "model-00001-of-00002.safetensors",
|
| 579 |
+
"audio_tower.layers.27.fc1.weight": "model-00001-of-00002.safetensors",
|
| 580 |
+
"audio_tower.layers.27.fc2.bias": "model-00001-of-00002.safetensors",
|
| 581 |
+
"audio_tower.layers.27.fc2.weight": "model-00001-of-00002.safetensors",
|
| 582 |
+
"audio_tower.layers.27.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 583 |
+
"audio_tower.layers.27.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 584 |
+
"audio_tower.layers.27.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 585 |
+
"audio_tower.layers.27.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 586 |
+
"audio_tower.layers.27.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 587 |
+
"audio_tower.layers.27.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 588 |
+
"audio_tower.layers.27.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 589 |
+
"audio_tower.layers.27.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 590 |
+
"audio_tower.layers.27.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 591 |
+
"audio_tower.layers.27.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 592 |
+
"audio_tower.layers.27.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 593 |
+
"audio_tower.layers.28.fc1.bias": "model-00001-of-00002.safetensors",
|
| 594 |
+
"audio_tower.layers.28.fc1.weight": "model-00001-of-00002.safetensors",
|
| 595 |
+
"audio_tower.layers.28.fc2.bias": "model-00001-of-00002.safetensors",
|
| 596 |
+
"audio_tower.layers.28.fc2.weight": "model-00001-of-00002.safetensors",
|
| 597 |
+
"audio_tower.layers.28.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 598 |
+
"audio_tower.layers.28.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 599 |
+
"audio_tower.layers.28.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 600 |
+
"audio_tower.layers.28.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 601 |
+
"audio_tower.layers.28.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 602 |
+
"audio_tower.layers.28.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 603 |
+
"audio_tower.layers.28.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 604 |
+
"audio_tower.layers.28.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 605 |
+
"audio_tower.layers.28.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 606 |
+
"audio_tower.layers.28.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 607 |
+
"audio_tower.layers.28.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 608 |
+
"audio_tower.layers.29.fc1.bias": "model-00001-of-00002.safetensors",
|
| 609 |
+
"audio_tower.layers.29.fc1.weight": "model-00001-of-00002.safetensors",
|
| 610 |
+
"audio_tower.layers.29.fc2.bias": "model-00001-of-00002.safetensors",
|
| 611 |
+
"audio_tower.layers.29.fc2.weight": "model-00001-of-00002.safetensors",
|
| 612 |
+
"audio_tower.layers.29.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 613 |
+
"audio_tower.layers.29.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 614 |
+
"audio_tower.layers.29.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 615 |
+
"audio_tower.layers.29.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 616 |
+
"audio_tower.layers.29.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 617 |
+
"audio_tower.layers.29.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 618 |
+
"audio_tower.layers.29.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 619 |
+
"audio_tower.layers.29.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 620 |
+
"audio_tower.layers.29.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 621 |
+
"audio_tower.layers.29.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 622 |
+
"audio_tower.layers.29.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 623 |
+
"audio_tower.layers.3.fc1.bias": "model-00001-of-00002.safetensors",
|
| 624 |
+
"audio_tower.layers.3.fc1.weight": "model-00001-of-00002.safetensors",
|
| 625 |
+
"audio_tower.layers.3.fc2.bias": "model-00001-of-00002.safetensors",
|
| 626 |
+
"audio_tower.layers.3.fc2.weight": "model-00001-of-00002.safetensors",
|
| 627 |
+
"audio_tower.layers.3.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 628 |
+
"audio_tower.layers.3.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 629 |
+
"audio_tower.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 630 |
+
"audio_tower.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 631 |
+
"audio_tower.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 632 |
+
"audio_tower.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 633 |
+
"audio_tower.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 634 |
+
"audio_tower.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 635 |
+
"audio_tower.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 636 |
+
"audio_tower.layers.3.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 637 |
+
"audio_tower.layers.3.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 638 |
+
"audio_tower.layers.30.fc1.bias": "model-00001-of-00002.safetensors",
|
| 639 |
+
"audio_tower.layers.30.fc1.weight": "model-00001-of-00002.safetensors",
|
| 640 |
+
"audio_tower.layers.30.fc2.bias": "model-00001-of-00002.safetensors",
|
| 641 |
+
"audio_tower.layers.30.fc2.weight": "model-00001-of-00002.safetensors",
|
| 642 |
+
"audio_tower.layers.30.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 643 |
+
"audio_tower.layers.30.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 644 |
+
"audio_tower.layers.30.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 645 |
+
"audio_tower.layers.30.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 646 |
+
"audio_tower.layers.30.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 647 |
+
"audio_tower.layers.30.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 648 |
+
"audio_tower.layers.30.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 649 |
+
"audio_tower.layers.30.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 650 |
+
"audio_tower.layers.30.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 651 |
+
"audio_tower.layers.30.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 652 |
+
"audio_tower.layers.30.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 653 |
+
"audio_tower.layers.31.fc1.bias": "model-00001-of-00002.safetensors",
|
| 654 |
+
"audio_tower.layers.31.fc1.weight": "model-00001-of-00002.safetensors",
|
| 655 |
+
"audio_tower.layers.31.fc2.bias": "model-00001-of-00002.safetensors",
|
| 656 |
+
"audio_tower.layers.31.fc2.weight": "model-00001-of-00002.safetensors",
|
| 657 |
+
"audio_tower.layers.31.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 658 |
+
"audio_tower.layers.31.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 659 |
+
"audio_tower.layers.31.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 660 |
+
"audio_tower.layers.31.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 661 |
+
"audio_tower.layers.31.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 662 |
+
"audio_tower.layers.31.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 663 |
+
"audio_tower.layers.31.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 664 |
+
"audio_tower.layers.31.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 665 |
+
"audio_tower.layers.31.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 666 |
+
"audio_tower.layers.31.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 667 |
+
"audio_tower.layers.31.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 668 |
+
"audio_tower.layers.4.fc1.bias": "model-00001-of-00002.safetensors",
|
| 669 |
+
"audio_tower.layers.4.fc1.weight": "model-00001-of-00002.safetensors",
|
| 670 |
+
"audio_tower.layers.4.fc2.bias": "model-00001-of-00002.safetensors",
|
| 671 |
+
"audio_tower.layers.4.fc2.weight": "model-00001-of-00002.safetensors",
|
| 672 |
+
"audio_tower.layers.4.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 673 |
+
"audio_tower.layers.4.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 674 |
+
"audio_tower.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 675 |
+
"audio_tower.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 676 |
+
"audio_tower.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 677 |
+
"audio_tower.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 678 |
+
"audio_tower.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 679 |
+
"audio_tower.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 680 |
+
"audio_tower.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 681 |
+
"audio_tower.layers.4.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 682 |
+
"audio_tower.layers.4.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 683 |
+
"audio_tower.layers.5.fc1.bias": "model-00001-of-00002.safetensors",
|
| 684 |
+
"audio_tower.layers.5.fc1.weight": "model-00001-of-00002.safetensors",
|
| 685 |
+
"audio_tower.layers.5.fc2.bias": "model-00001-of-00002.safetensors",
|
| 686 |
+
"audio_tower.layers.5.fc2.weight": "model-00001-of-00002.safetensors",
|
| 687 |
+
"audio_tower.layers.5.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 688 |
+
"audio_tower.layers.5.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 689 |
+
"audio_tower.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 690 |
+
"audio_tower.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 691 |
+
"audio_tower.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 692 |
+
"audio_tower.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 693 |
+
"audio_tower.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 694 |
+
"audio_tower.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 695 |
+
"audio_tower.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 696 |
+
"audio_tower.layers.5.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 697 |
+
"audio_tower.layers.5.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 698 |
+
"audio_tower.layers.6.fc1.bias": "model-00001-of-00002.safetensors",
|
| 699 |
+
"audio_tower.layers.6.fc1.weight": "model-00001-of-00002.safetensors",
|
| 700 |
+
"audio_tower.layers.6.fc2.bias": "model-00001-of-00002.safetensors",
|
| 701 |
+
"audio_tower.layers.6.fc2.weight": "model-00001-of-00002.safetensors",
|
| 702 |
+
"audio_tower.layers.6.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 703 |
+
"audio_tower.layers.6.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 704 |
+
"audio_tower.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 705 |
+
"audio_tower.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 706 |
+
"audio_tower.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 707 |
+
"audio_tower.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 708 |
+
"audio_tower.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 709 |
+
"audio_tower.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 710 |
+
"audio_tower.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 711 |
+
"audio_tower.layers.6.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 712 |
+
"audio_tower.layers.6.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 713 |
+
"audio_tower.layers.7.fc1.bias": "model-00001-of-00002.safetensors",
|
| 714 |
+
"audio_tower.layers.7.fc1.weight": "model-00001-of-00002.safetensors",
|
| 715 |
+
"audio_tower.layers.7.fc2.bias": "model-00001-of-00002.safetensors",
|
| 716 |
+
"audio_tower.layers.7.fc2.weight": "model-00001-of-00002.safetensors",
|
| 717 |
+
"audio_tower.layers.7.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 718 |
+
"audio_tower.layers.7.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 719 |
+
"audio_tower.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 720 |
+
"audio_tower.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 721 |
+
"audio_tower.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 722 |
+
"audio_tower.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 723 |
+
"audio_tower.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 724 |
+
"audio_tower.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 725 |
+
"audio_tower.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 726 |
+
"audio_tower.layers.7.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 727 |
+
"audio_tower.layers.7.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 728 |
+
"audio_tower.layers.8.fc1.bias": "model-00001-of-00002.safetensors",
|
| 729 |
+
"audio_tower.layers.8.fc1.weight": "model-00001-of-00002.safetensors",
|
| 730 |
+
"audio_tower.layers.8.fc2.bias": "model-00001-of-00002.safetensors",
|
| 731 |
+
"audio_tower.layers.8.fc2.weight": "model-00001-of-00002.safetensors",
|
| 732 |
+
"audio_tower.layers.8.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 733 |
+
"audio_tower.layers.8.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 734 |
+
"audio_tower.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 735 |
+
"audio_tower.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 736 |
+
"audio_tower.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 737 |
+
"audio_tower.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 738 |
+
"audio_tower.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 739 |
+
"audio_tower.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 740 |
+
"audio_tower.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 741 |
+
"audio_tower.layers.8.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 742 |
+
"audio_tower.layers.8.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 743 |
+
"audio_tower.layers.9.fc1.bias": "model-00001-of-00002.safetensors",
|
| 744 |
+
"audio_tower.layers.9.fc1.weight": "model-00001-of-00002.safetensors",
|
| 745 |
+
"audio_tower.layers.9.fc2.bias": "model-00001-of-00002.safetensors",
|
| 746 |
+
"audio_tower.layers.9.fc2.weight": "model-00001-of-00002.safetensors",
|
| 747 |
+
"audio_tower.layers.9.final_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 748 |
+
"audio_tower.layers.9.final_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 749 |
+
"audio_tower.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 750 |
+
"audio_tower.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors",
|
| 751 |
+
"audio_tower.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors",
|
| 752 |
+
"audio_tower.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
|
| 753 |
+
"audio_tower.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 754 |
+
"audio_tower.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
|
| 755 |
+
"audio_tower.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 756 |
+
"audio_tower.layers.9.self_attn_layer_norm.bias": "model-00001-of-00002.safetensors",
|
| 757 |
+
"audio_tower.layers.9.self_attn_layer_norm.weight": "model-00001-of-00002.safetensors",
|
| 758 |
+
"language_model.lm_head.weight": "model-00002-of-00002.safetensors",
|
| 759 |
+
"language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors",
|
| 760 |
+
"language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 761 |
+
"language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 762 |
+
"language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 763 |
+
"language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 764 |
+
"language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 765 |
+
"language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 766 |
+
"language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 767 |
+
"language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 768 |
+
"language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 769 |
+
"language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 770 |
+
"language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 771 |
+
"language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 772 |
+
"language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 773 |
+
"language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 774 |
+
"language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 775 |
+
"language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 776 |
+
"language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 777 |
+
"language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 778 |
+
"language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 779 |
+
"language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 780 |
+
"language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 781 |
+
"language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 782 |
+
"language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 783 |
+
"language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 784 |
+
"language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 785 |
+
"language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 786 |
+
"language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 787 |
+
"language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 788 |
+
"language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 789 |
+
"language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 790 |
+
"language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 791 |
+
"language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 792 |
+
"language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 793 |
+
"language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 794 |
+
"language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 795 |
+
"language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 796 |
+
"language_model.model.layers.12.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 797 |
+
"language_model.model.layers.12.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 798 |
+
"language_model.model.layers.12.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 799 |
+
"language_model.model.layers.12.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 800 |
+
"language_model.model.layers.12.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 801 |
+
"language_model.model.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 802 |
+
"language_model.model.layers.12.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 803 |
+
"language_model.model.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 804 |
+
"language_model.model.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 805 |
+
"language_model.model.layers.13.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 806 |
+
"language_model.model.layers.13.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 807 |
+
"language_model.model.layers.13.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 808 |
+
"language_model.model.layers.13.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 809 |
+
"language_model.model.layers.13.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 810 |
+
"language_model.model.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 811 |
+
"language_model.model.layers.13.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 812 |
+
"language_model.model.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 813 |
+
"language_model.model.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 814 |
+
"language_model.model.layers.14.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 815 |
+
"language_model.model.layers.14.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 816 |
+
"language_model.model.layers.14.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 817 |
+
"language_model.model.layers.14.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 818 |
+
"language_model.model.layers.14.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 819 |
+
"language_model.model.layers.14.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 820 |
+
"language_model.model.layers.14.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 821 |
+
"language_model.model.layers.14.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 822 |
+
"language_model.model.layers.14.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 823 |
+
"language_model.model.layers.15.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 824 |
+
"language_model.model.layers.15.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 825 |
+
"language_model.model.layers.15.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 826 |
+
"language_model.model.layers.15.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 827 |
+
"language_model.model.layers.15.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 828 |
+
"language_model.model.layers.15.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 829 |
+
"language_model.model.layers.15.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 830 |
+
"language_model.model.layers.15.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 831 |
+
"language_model.model.layers.15.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 832 |
+
"language_model.model.layers.16.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 833 |
+
"language_model.model.layers.16.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 834 |
+
"language_model.model.layers.16.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 835 |
+
"language_model.model.layers.16.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 836 |
+
"language_model.model.layers.16.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 837 |
+
"language_model.model.layers.16.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 838 |
+
"language_model.model.layers.16.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 839 |
+
"language_model.model.layers.16.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 840 |
+
"language_model.model.layers.16.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 841 |
+
"language_model.model.layers.17.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 842 |
+
"language_model.model.layers.17.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 843 |
+
"language_model.model.layers.17.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 844 |
+
"language_model.model.layers.17.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 845 |
+
"language_model.model.layers.17.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 846 |
+
"language_model.model.layers.17.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 847 |
+
"language_model.model.layers.17.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 848 |
+
"language_model.model.layers.17.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 849 |
+
"language_model.model.layers.17.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 850 |
+
"language_model.model.layers.18.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 851 |
+
"language_model.model.layers.18.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 852 |
+
"language_model.model.layers.18.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 853 |
+
"language_model.model.layers.18.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 854 |
+
"language_model.model.layers.18.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 855 |
+
"language_model.model.layers.18.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 856 |
+
"language_model.model.layers.18.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 857 |
+
"language_model.model.layers.18.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 858 |
+
"language_model.model.layers.18.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 859 |
+
"language_model.model.layers.19.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 860 |
+
"language_model.model.layers.19.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 861 |
+
"language_model.model.layers.19.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 862 |
+
"language_model.model.layers.19.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 863 |
+
"language_model.model.layers.19.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 864 |
+
"language_model.model.layers.19.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 865 |
+
"language_model.model.layers.19.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 866 |
+
"language_model.model.layers.19.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 867 |
+
"language_model.model.layers.19.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 868 |
+
"language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 869 |
+
"language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 870 |
+
"language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 871 |
+
"language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 872 |
+
"language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 873 |
+
"language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 874 |
+
"language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 875 |
+
"language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 876 |
+
"language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 877 |
+
"language_model.model.layers.20.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 878 |
+
"language_model.model.layers.20.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 879 |
+
"language_model.model.layers.20.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 880 |
+
"language_model.model.layers.20.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 881 |
+
"language_model.model.layers.20.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 882 |
+
"language_model.model.layers.20.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 883 |
+
"language_model.model.layers.20.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 884 |
+
"language_model.model.layers.20.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 885 |
+
"language_model.model.layers.20.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 886 |
+
"language_model.model.layers.21.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 887 |
+
"language_model.model.layers.21.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 888 |
+
"language_model.model.layers.21.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 889 |
+
"language_model.model.layers.21.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 890 |
+
"language_model.model.layers.21.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 891 |
+
"language_model.model.layers.21.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 892 |
+
"language_model.model.layers.21.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 893 |
+
"language_model.model.layers.21.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 894 |
+
"language_model.model.layers.21.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 895 |
+
"language_model.model.layers.22.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 896 |
+
"language_model.model.layers.22.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 897 |
+
"language_model.model.layers.22.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 898 |
+
"language_model.model.layers.22.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 899 |
+
"language_model.model.layers.22.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 900 |
+
"language_model.model.layers.22.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 901 |
+
"language_model.model.layers.22.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 902 |
+
"language_model.model.layers.22.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 903 |
+
"language_model.model.layers.22.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 904 |
+
"language_model.model.layers.23.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 905 |
+
"language_model.model.layers.23.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 906 |
+
"language_model.model.layers.23.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 907 |
+
"language_model.model.layers.23.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 908 |
+
"language_model.model.layers.23.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 909 |
+
"language_model.model.layers.23.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 910 |
+
"language_model.model.layers.23.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 911 |
+
"language_model.model.layers.23.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 912 |
+
"language_model.model.layers.23.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 913 |
+
"language_model.model.layers.24.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 914 |
+
"language_model.model.layers.24.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 915 |
+
"language_model.model.layers.24.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 916 |
+
"language_model.model.layers.24.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 917 |
+
"language_model.model.layers.24.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 918 |
+
"language_model.model.layers.24.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 919 |
+
"language_model.model.layers.24.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 920 |
+
"language_model.model.layers.24.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 921 |
+
"language_model.model.layers.24.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 922 |
+
"language_model.model.layers.25.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 923 |
+
"language_model.model.layers.25.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 924 |
+
"language_model.model.layers.25.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 925 |
+
"language_model.model.layers.25.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 926 |
+
"language_model.model.layers.25.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 927 |
+
"language_model.model.layers.25.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 928 |
+
"language_model.model.layers.25.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 929 |
+
"language_model.model.layers.25.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 930 |
+
"language_model.model.layers.25.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 931 |
+
"language_model.model.layers.26.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 932 |
+
"language_model.model.layers.26.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 933 |
+
"language_model.model.layers.26.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 934 |
+
"language_model.model.layers.26.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 935 |
+
"language_model.model.layers.26.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 936 |
+
"language_model.model.layers.26.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 937 |
+
"language_model.model.layers.26.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 938 |
+
"language_model.model.layers.26.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 939 |
+
"language_model.model.layers.26.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 940 |
+
"language_model.model.layers.27.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 941 |
+
"language_model.model.layers.27.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 942 |
+
"language_model.model.layers.27.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 943 |
+
"language_model.model.layers.27.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 944 |
+
"language_model.model.layers.27.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 945 |
+
"language_model.model.layers.27.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 946 |
+
"language_model.model.layers.27.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 947 |
+
"language_model.model.layers.27.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 948 |
+
"language_model.model.layers.27.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 949 |
+
"language_model.model.layers.28.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 950 |
+
"language_model.model.layers.28.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 951 |
+
"language_model.model.layers.28.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 952 |
+
"language_model.model.layers.28.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 953 |
+
"language_model.model.layers.28.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 954 |
+
"language_model.model.layers.28.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 955 |
+
"language_model.model.layers.28.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 956 |
+
"language_model.model.layers.28.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 957 |
+
"language_model.model.layers.28.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 958 |
+
"language_model.model.layers.29.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 959 |
+
"language_model.model.layers.29.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 960 |
+
"language_model.model.layers.29.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 961 |
+
"language_model.model.layers.29.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 962 |
+
"language_model.model.layers.29.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 963 |
+
"language_model.model.layers.29.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 964 |
+
"language_model.model.layers.29.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 965 |
+
"language_model.model.layers.29.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 966 |
+
"language_model.model.layers.29.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 967 |
+
"language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 968 |
+
"language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 969 |
+
"language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 970 |
+
"language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 971 |
+
"language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 972 |
+
"language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 973 |
+
"language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 974 |
+
"language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 975 |
+
"language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 976 |
+
"language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 977 |
+
"language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 978 |
+
"language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 979 |
+
"language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 980 |
+
"language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 981 |
+
"language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 982 |
+
"language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 983 |
+
"language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 984 |
+
"language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 985 |
+
"language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 986 |
+
"language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 987 |
+
"language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 988 |
+
"language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 989 |
+
"language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 990 |
+
"language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 991 |
+
"language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 992 |
+
"language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 993 |
+
"language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 994 |
+
"language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 995 |
+
"language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 996 |
+
"language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 997 |
+
"language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 998 |
+
"language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 999 |
+
"language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 1000 |
+
"language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 1001 |
+
"language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 1002 |
+
"language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 1003 |
+
"language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 1004 |
+
"language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 1005 |
+
"language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 1006 |
+
"language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 1007 |
+
"language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 1008 |
+
"language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 1009 |
+
"language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 1010 |
+
"language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 1011 |
+
"language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 1012 |
+
"language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 1013 |
+
"language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 1014 |
+
"language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 1015 |
+
"language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 1016 |
+
"language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 1017 |
+
"language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 1018 |
+
"language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 1019 |
+
"language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 1020 |
+
"language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 1021 |
+
"language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 1022 |
+
"language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 1023 |
+
"language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 1024 |
+
"language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 1025 |
+
"language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 1026 |
+
"language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 1027 |
+
"language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 1028 |
+
"language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 1029 |
+
"language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 1030 |
+
"language_model.model.norm.weight": "model-00002-of-00002.safetensors",
|
| 1031 |
+
"multi_modal_projector.linear_1.weight": "model-00002-of-00002.safetensors",
|
| 1032 |
+
"multi_modal_projector.linear_2.weight": "model-00002-of-00002.safetensors"
|
| 1033 |
+
}
|
| 1034 |
+
}
|
modeling_dixtral.py
ADDED
|
@@ -0,0 +1,923 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# coding=utf-8
|
| 2 |
+
# Copyright 2025 The HuggingFace Inc. team. All rights reserved.
|
| 3 |
+
#
|
| 4 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
| 5 |
+
# you may not use this file except in compliance with the License.
|
| 6 |
+
# You may obtain a copy of the License at
|
| 7 |
+
#
|
| 8 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
| 9 |
+
#
|
| 10 |
+
# Unless required by applicable law or agreed to in writing, software
|
| 11 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
| 12 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 13 |
+
# See the License for the specific language governing permissions and
|
| 14 |
+
# limitations under the License.
|
| 15 |
+
import copy
|
| 16 |
+
import math
|
| 17 |
+
from typing import Callable, Optional, Union, Any, Dict
|
| 18 |
+
|
| 19 |
+
import wandb
|
| 20 |
+
import torch
|
| 21 |
+
from torch import nn
|
| 22 |
+
from transformers.activations import ACT2FN
|
| 23 |
+
from transformers.cache_utils import Cache
|
| 24 |
+
from transformers.generation import GenerationMixin
|
| 25 |
+
from transformers.modeling_layers import GradientCheckpointingLayer
|
| 26 |
+
from transformers.modeling_outputs import BaseModelOutput, BaseModelOutputWithPast, CausalLMOutputWithPast
|
| 27 |
+
from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
|
| 28 |
+
from transformers.processing_utils import Unpack
|
| 29 |
+
from transformers.utils import TransformersKwargs, auto_docstring, can_return_tuple, logging
|
| 30 |
+
from transformers.utils.generic import check_model_inputs
|
| 31 |
+
from transformers.models.auto import AutoModel, AutoModelForCausalLM
|
| 32 |
+
from .configuration_dixtral import DixtralConfig, DixtralEncoderConfig
|
| 33 |
+
from transformers.models.voxtral import VoxtralConfig
|
| 34 |
+
from transformers.generation.utils import GenerationConfig, LogitsProcessorList
|
| 35 |
+
from src.models.dicow.FDDT import FDDT
|
| 36 |
+
from src.models.dicow.layers import CustomLinear, CustomDiagonalLinear
|
| 37 |
+
from src.models.dixtral.decoding import CTCRescorerLogitsProcessorWithPruning
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
logger = logging.get_logger(__name__)
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
def eager_attention_forward(
|
| 44 |
+
module: nn.Module,
|
| 45 |
+
query: torch.Tensor,
|
| 46 |
+
key: torch.Tensor,
|
| 47 |
+
value: torch.Tensor,
|
| 48 |
+
attention_mask: Optional[torch.Tensor],
|
| 49 |
+
scaling: Optional[float] = None,
|
| 50 |
+
dropout: float = 0.0,
|
| 51 |
+
head_mask: Optional[torch.Tensor] = None,
|
| 52 |
+
**kwargs,
|
| 53 |
+
):
|
| 54 |
+
if scaling is None:
|
| 55 |
+
scaling = query.size(-1) ** -0.5
|
| 56 |
+
|
| 57 |
+
attn_weights = torch.matmul(query, key.transpose(2, 3)) * scaling
|
| 58 |
+
if attention_mask is not None and attention_mask.ndim == 4:
|
| 59 |
+
attn_weights = attn_weights + attention_mask[:, :, :, : key.shape[-2]]
|
| 60 |
+
|
| 61 |
+
attn_weights = nn.functional.softmax(attn_weights, dim=-1)
|
| 62 |
+
|
| 63 |
+
if head_mask is not None:
|
| 64 |
+
attn_weights = attn_weights * head_mask.view(1, -1, 1, 1)
|
| 65 |
+
|
| 66 |
+
attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
|
| 67 |
+
attn_output = torch.matmul(attn_weights, value)
|
| 68 |
+
attn_output = attn_output.transpose(1, 2).contiguous()
|
| 69 |
+
|
| 70 |
+
return attn_output, attn_weights
|
| 71 |
+
|
| 72 |
+
|
| 73 |
+
class CTCProcessorDummy:
|
| 74 |
+
def __init__(self):
|
| 75 |
+
super().__init__()
|
| 76 |
+
self.func = None
|
| 77 |
+
def set_func(self,func):
|
| 78 |
+
self.func = func
|
| 79 |
+
def __call__(self, input_ids_orig: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor:
|
| 80 |
+
return self.func(input_ids_orig, scores)
|
| 81 |
+
|
| 82 |
+
class VoxtralAttention(nn.Module):
|
| 83 |
+
"""Multi-headed attention from 'Attention Is All You Need' paper"""
|
| 84 |
+
|
| 85 |
+
def __init__(
|
| 86 |
+
self,
|
| 87 |
+
embed_dim: int,
|
| 88 |
+
num_heads: int,
|
| 89 |
+
dropout: float = 0.0,
|
| 90 |
+
is_decoder: bool = False,
|
| 91 |
+
bias: bool = True,
|
| 92 |
+
is_causal: bool = False,
|
| 93 |
+
layer_idx: Optional[int] = None,
|
| 94 |
+
config: Optional[VoxtralConfig] = None,
|
| 95 |
+
):
|
| 96 |
+
super().__init__()
|
| 97 |
+
self.embed_dim = embed_dim
|
| 98 |
+
self.num_heads = num_heads
|
| 99 |
+
self.dropout = dropout
|
| 100 |
+
self.head_dim = embed_dim // num_heads
|
| 101 |
+
self.config = config
|
| 102 |
+
|
| 103 |
+
if (self.head_dim * num_heads) != self.embed_dim:
|
| 104 |
+
raise ValueError(
|
| 105 |
+
f"embed_dim must be divisible by num_heads (got `embed_dim`: {self.embed_dim}"
|
| 106 |
+
f" and `num_heads`: {num_heads})."
|
| 107 |
+
)
|
| 108 |
+
self.scaling = self.head_dim**-0.5
|
| 109 |
+
self.is_decoder = is_decoder
|
| 110 |
+
self.is_causal = is_causal
|
| 111 |
+
|
| 112 |
+
if layer_idx is None and is_decoder:
|
| 113 |
+
logger.warning_once(
|
| 114 |
+
f"Instantiating a decoder {self.__class__.__name__} without passing `layer_idx` is not recommended and "
|
| 115 |
+
"will to errors during the forward call, if caching is used. Please make sure to provide a `layer_idx` "
|
| 116 |
+
"when creating this class."
|
| 117 |
+
)
|
| 118 |
+
self.layer_idx = layer_idx
|
| 119 |
+
|
| 120 |
+
self.k_proj = nn.Linear(embed_dim, embed_dim, bias=False)
|
| 121 |
+
self.v_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
|
| 122 |
+
self.q_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
|
| 123 |
+
self.out_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
|
| 124 |
+
|
| 125 |
+
def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
|
| 126 |
+
return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
|
| 127 |
+
|
| 128 |
+
def forward(
|
| 129 |
+
self,
|
| 130 |
+
hidden_states: torch.Tensor,
|
| 131 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 132 |
+
layer_head_mask: Optional[torch.Tensor] = None,
|
| 133 |
+
output_attentions: bool = False,
|
| 134 |
+
**kwargs,
|
| 135 |
+
) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
|
| 136 |
+
"""Input shape: Batch x Time x Channel"""
|
| 137 |
+
|
| 138 |
+
bsz, tgt_len, _ = hidden_states.size()
|
| 139 |
+
|
| 140 |
+
# Scaling is susceptible to floating point arithmetics' inprecisions
|
| 141 |
+
# which can lead to different results (this is dependent from model
|
| 142 |
+
# to model, e.g. whisper is one such case). We therefore keep the
|
| 143 |
+
# original order of scaling to follow the original implementation
|
| 144 |
+
# and enforce no scaling (1.0) in the attention call below.
|
| 145 |
+
query_states = self._shape(self.q_proj(hidden_states) * self.scaling, tgt_len, bsz)
|
| 146 |
+
key_states = self._shape(self.k_proj(hidden_states), -1, bsz)
|
| 147 |
+
value_states = self._shape(self.v_proj(hidden_states), -1, bsz)
|
| 148 |
+
|
| 149 |
+
attention_interface: Callable = eager_attention_forward
|
| 150 |
+
if self.config._attn_implementation != "eager":
|
| 151 |
+
attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
|
| 152 |
+
|
| 153 |
+
attn_output, attn_weights = attention_interface(
|
| 154 |
+
self,
|
| 155 |
+
query_states,
|
| 156 |
+
key_states,
|
| 157 |
+
value_states,
|
| 158 |
+
attention_mask,
|
| 159 |
+
dropout=0.0 if not self.training else self.dropout,
|
| 160 |
+
scaling=1.0,
|
| 161 |
+
output_attentions=output_attentions,
|
| 162 |
+
head_mask=layer_head_mask,
|
| 163 |
+
**kwargs,
|
| 164 |
+
)
|
| 165 |
+
|
| 166 |
+
attn_output = attn_output.reshape(bsz, tgt_len, -1).contiguous()
|
| 167 |
+
attn_output = self.out_proj(attn_output)
|
| 168 |
+
|
| 169 |
+
return attn_output, attn_weights
|
| 170 |
+
|
| 171 |
+
|
| 172 |
+
class VoxtralEncoderLayer(GradientCheckpointingLayer):
|
| 173 |
+
def __init__(self, config: VoxtralConfig):
|
| 174 |
+
super().__init__()
|
| 175 |
+
self.embed_dim = config.d_model
|
| 176 |
+
|
| 177 |
+
self.self_attn = VoxtralAttention(
|
| 178 |
+
embed_dim=self.embed_dim,
|
| 179 |
+
num_heads=config.encoder_attention_heads,
|
| 180 |
+
dropout=config.attention_dropout,
|
| 181 |
+
config=config,
|
| 182 |
+
)
|
| 183 |
+
self.self_attn_layer_norm = nn.LayerNorm(self.embed_dim)
|
| 184 |
+
self.dropout = config.dropout
|
| 185 |
+
self.activation_fn = ACT2FN[config.activation_function]
|
| 186 |
+
self.activation_dropout = config.activation_dropout
|
| 187 |
+
self.fc1 = nn.Linear(self.embed_dim, config.encoder_ffn_dim)
|
| 188 |
+
self.fc2 = nn.Linear(config.encoder_ffn_dim, self.embed_dim)
|
| 189 |
+
self.final_layer_norm = nn.LayerNorm(self.embed_dim)
|
| 190 |
+
|
| 191 |
+
def forward(
|
| 192 |
+
self,
|
| 193 |
+
hidden_states: torch.Tensor,
|
| 194 |
+
attention_mask: torch.Tensor,
|
| 195 |
+
layer_head_mask: torch.Tensor,
|
| 196 |
+
output_attentions: bool = False,
|
| 197 |
+
) -> torch.Tensor:
|
| 198 |
+
"""
|
| 199 |
+
Args:
|
| 200 |
+
hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
|
| 201 |
+
attention_mask (`torch.FloatTensor`): attention mask of size
|
| 202 |
+
`(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
|
| 203 |
+
layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size
|
| 204 |
+
`(encoder_attention_heads,)`.
|
| 205 |
+
output_attentions (`bool`, *optional*):
|
| 206 |
+
Whether or not to return the attentions tensors of all attention layers. See `attentions` under
|
| 207 |
+
returned tensors for more detail.
|
| 208 |
+
"""
|
| 209 |
+
residual = hidden_states
|
| 210 |
+
hidden_states = self.self_attn_layer_norm(hidden_states)
|
| 211 |
+
hidden_states, attn_weights = self.self_attn(
|
| 212 |
+
hidden_states=hidden_states,
|
| 213 |
+
attention_mask=attention_mask,
|
| 214 |
+
layer_head_mask=layer_head_mask,
|
| 215 |
+
output_attentions=output_attentions,
|
| 216 |
+
)
|
| 217 |
+
hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
|
| 218 |
+
hidden_states = residual + hidden_states
|
| 219 |
+
|
| 220 |
+
residual = hidden_states
|
| 221 |
+
hidden_states = self.final_layer_norm(hidden_states)
|
| 222 |
+
hidden_states = self.activation_fn(self.fc1(hidden_states))
|
| 223 |
+
hidden_states = nn.functional.dropout(hidden_states, p=self.activation_dropout, training=self.training)
|
| 224 |
+
hidden_states = self.fc2(hidden_states)
|
| 225 |
+
hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
|
| 226 |
+
hidden_states = residual + hidden_states
|
| 227 |
+
|
| 228 |
+
if hidden_states.dtype == torch.float16:
|
| 229 |
+
clamp_value = torch.finfo(hidden_states.dtype).max - 1000
|
| 230 |
+
hidden_states = torch.clamp(hidden_states, min=-clamp_value, max=clamp_value)
|
| 231 |
+
|
| 232 |
+
return hidden_states, attn_weights
|
| 233 |
+
|
| 234 |
+
|
| 235 |
+
@auto_docstring
|
| 236 |
+
class DixtralPreTrainedModel(PreTrainedModel):
|
| 237 |
+
config: DixtralConfig
|
| 238 |
+
base_model_prefix = "model"
|
| 239 |
+
supports_gradient_checkpointing = True
|
| 240 |
+
_no_split_modules = None
|
| 241 |
+
_skip_keys_device_placement = "past_key_values"
|
| 242 |
+
_supports_flash_attn = True
|
| 243 |
+
_supports_sdpa = True
|
| 244 |
+
_supports_flex_attn = True
|
| 245 |
+
_supports_cache_class = True
|
| 246 |
+
_supports_attention_backend = True
|
| 247 |
+
_can_compile_fullgraph = True
|
| 248 |
+
|
| 249 |
+
def _init_weights(self, module):
|
| 250 |
+
# important: this ported version of Voxtral isn't meant for training from scratch - only
|
| 251 |
+
# inference and fine-tuning - so the proper init weights code has been removed
|
| 252 |
+
std = (
|
| 253 |
+
self.config.initializer_range
|
| 254 |
+
if hasattr(self.config, "initializer_range")
|
| 255 |
+
else self.config.audio_config.initializer_range
|
| 256 |
+
)
|
| 257 |
+
|
| 258 |
+
if isinstance(module, (nn.Linear, nn.Conv1d)):
|
| 259 |
+
module.weight.data.normal_(mean=0.0, std=std)
|
| 260 |
+
if module.bias is not None:
|
| 261 |
+
module.bias.data.zero_()
|
| 262 |
+
elif isinstance(module, nn.LayerNorm):
|
| 263 |
+
module.weight.data.fill_(1.0)
|
| 264 |
+
module.bias.data.zero_()
|
| 265 |
+
elif isinstance(module, nn.Embedding):
|
| 266 |
+
module.weight.data.normal_(mean=0.0, std=std)
|
| 267 |
+
if module.padding_idx is not None:
|
| 268 |
+
module.weight.data[module.padding_idx].zero_()
|
| 269 |
+
elif isinstance(module, (CustomLinear, CustomDiagonalLinear)):
|
| 270 |
+
module.reset_parameters()
|
| 271 |
+
|
| 272 |
+
|
| 273 |
+
@auto_docstring(
|
| 274 |
+
custom_intro="""
|
| 275 |
+
The Voxtral encoder, which is a Whisper encoder.
|
| 276 |
+
"""
|
| 277 |
+
)
|
| 278 |
+
class DixtralEncoder(DixtralPreTrainedModel):
|
| 279 |
+
"""
|
| 280 |
+
Transformer encoder consisting of *config.encoder_layers* self attention layers. Each layer is a
|
| 281 |
+
[`VoxtralEncoderLayer`].
|
| 282 |
+
|
| 283 |
+
Args:
|
| 284 |
+
config: VoxtralEncoderConfig
|
| 285 |
+
"""
|
| 286 |
+
|
| 287 |
+
# Ignore copy
|
| 288 |
+
config: DixtralEncoderConfig
|
| 289 |
+
main_input_name = "input_features"
|
| 290 |
+
_no_split_modules = ["VoxtralEncoderLayer"]
|
| 291 |
+
_can_record_outputs = {
|
| 292 |
+
"attentions": VoxtralAttention,
|
| 293 |
+
"hidden_states": VoxtralEncoderLayer,
|
| 294 |
+
}
|
| 295 |
+
|
| 296 |
+
def __init__(self, config: DixtralEncoderConfig):
|
| 297 |
+
super().__init__(config)
|
| 298 |
+
self.dropout = config.dropout
|
| 299 |
+
self.layerdrop = config.encoder_layerdrop
|
| 300 |
+
|
| 301 |
+
embed_dim = config.d_model
|
| 302 |
+
self.num_mel_bins = config.num_mel_bins
|
| 303 |
+
self.padding_idx = config.pad_token_id
|
| 304 |
+
self.max_source_positions = config.max_source_positions
|
| 305 |
+
self.embed_scale = math.sqrt(embed_dim) if config.scale_embedding else 1.0
|
| 306 |
+
|
| 307 |
+
self.conv1 = nn.Conv1d(self.num_mel_bins, embed_dim, kernel_size=3, padding=1)
|
| 308 |
+
self.conv2 = nn.Conv1d(embed_dim, embed_dim, kernel_size=3, stride=2, padding=1)
|
| 309 |
+
|
| 310 |
+
self.embed_positions = nn.Embedding(self.max_source_positions, embed_dim)
|
| 311 |
+
self.embed_positions.requires_grad_(False)
|
| 312 |
+
|
| 313 |
+
self.layers = nn.ModuleList([VoxtralEncoderLayer(config) for _ in range(config.encoder_layers)])
|
| 314 |
+
self.layer_norm = nn.LayerNorm(config.d_model)
|
| 315 |
+
# Ignore copy
|
| 316 |
+
self.avg_pooler = nn.AvgPool1d(2, stride=2)
|
| 317 |
+
|
| 318 |
+
self._init_dicow_components(config)
|
| 319 |
+
|
| 320 |
+
self.gradient_checkpointing = False
|
| 321 |
+
# Initialize weights and apply final processing
|
| 322 |
+
self.post_init()
|
| 323 |
+
|
| 324 |
+
def _init_dicow_components(self, config):
|
| 325 |
+
"""Initialize DiCoW-specific components"""
|
| 326 |
+
if not config.use_dicow_encoder:
|
| 327 |
+
return
|
| 328 |
+
|
| 329 |
+
# FDDT components
|
| 330 |
+
if config.use_fddt:
|
| 331 |
+
num_fddts = (config.apply_fddt_to_n_layers
|
| 332 |
+
if config.apply_fddt_to_n_layers != -1
|
| 333 |
+
else len(self.layers))
|
| 334 |
+
self.fddts = nn.ModuleList([
|
| 335 |
+
FDDT(
|
| 336 |
+
d_model=config.d_model,
|
| 337 |
+
non_target_rate=1.0,
|
| 338 |
+
fddt_init=config.fddt_init,
|
| 339 |
+
is_diagonal=config.fddt_is_diagonal,
|
| 340 |
+
bias_only=config.fddt_bias_only,
|
| 341 |
+
use_silence=config.fddt_use_silence,
|
| 342 |
+
use_target=config.fddt_use_target,
|
| 343 |
+
use_overlap=config.fddt_use_overlap,
|
| 344 |
+
use_non_target=config.fddt_use_non_target,
|
| 345 |
+
)
|
| 346 |
+
for _ in range(num_fddts)
|
| 347 |
+
])
|
| 348 |
+
|
| 349 |
+
if config.use_pre_pos_fddt:
|
| 350 |
+
self.initial_fddt = FDDT(
|
| 351 |
+
d_model=config.d_model,
|
| 352 |
+
non_target_rate=config.non_target_fddt_value,
|
| 353 |
+
fddt_init=config.fddt_init,
|
| 354 |
+
is_diagonal=config.fddt_is_diagonal,
|
| 355 |
+
bias_only=config.fddt_bias_only,
|
| 356 |
+
use_silence=config.fddt_use_silence,
|
| 357 |
+
use_target=config.fddt_use_target,
|
| 358 |
+
use_overlap=config.fddt_use_overlap,
|
| 359 |
+
use_non_target=config.fddt_use_non_target,
|
| 360 |
+
)
|
| 361 |
+
|
| 362 |
+
# For CTC label processing
|
| 363 |
+
self.first_task_token = config.vocab_size - 30 * 50 - 1 - 6
|
| 364 |
+
|
| 365 |
+
def _freeze_parameters(self):
|
| 366 |
+
for param in self.parameters():
|
| 367 |
+
param.requires_grad = False
|
| 368 |
+
self._requires_grad = False
|
| 369 |
+
|
| 370 |
+
def get_input_embeddings(self) -> nn.Module:
|
| 371 |
+
return self.conv1
|
| 372 |
+
|
| 373 |
+
def set_input_embeddings(self, value: nn.Module):
|
| 374 |
+
self.conv1 = value
|
| 375 |
+
|
| 376 |
+
|
| 377 |
+
@check_model_inputs
|
| 378 |
+
def forward(
|
| 379 |
+
self,
|
| 380 |
+
input_features,
|
| 381 |
+
attention_mask=None,
|
| 382 |
+
stno_mask=None,
|
| 383 |
+
**kwargs: Unpack[TransformersKwargs],
|
| 384 |
+
):
|
| 385 |
+
r"""
|
| 386 |
+
Args:
|
| 387 |
+
input_features (`torch.LongTensor` of shape `(batch_size, feature_size, sequence_length)`):
|
| 388 |
+
Float values of mel features extracted from the raw speech waveform. Raw speech waveform can be
|
| 389 |
+
obtained by loading a `.flac` or `.wav` audio file into an array of type `list[float]` or a
|
| 390 |
+
`numpy.ndarray`, *e.g.* via the soundfile library (`pip install soundfile`). To prepare the array into
|
| 391 |
+
`input_features`, the [`AutoFeatureExtractor`] should be used for extracting the mel features, padding
|
| 392 |
+
and conversion into a tensor of type `torch.FloatTensor`. See [`~WhisperFeatureExtractor.__call__`]
|
| 393 |
+
attention_mask (`torch.Tensor`)`, *optional*):
|
| 394 |
+
Voxtral does not support masking of the `input_features`, this argument is preserved for compatibility,
|
| 395 |
+
but it is not used. By default the silence in the input log mel spectrogram are ignored.
|
| 396 |
+
"""
|
| 397 |
+
expected_seq_length = self.config.max_source_positions * self.conv1.stride[0] * self.conv2.stride[0]
|
| 398 |
+
if input_features.shape[-1] != expected_seq_length:
|
| 399 |
+
raise ValueError(
|
| 400 |
+
f"Qwen2Audio expects the mel input features to be of length {expected_seq_length}, but found {input_features.shape[-1]}. Make sure to pad the input mel features to {expected_seq_length}."
|
| 401 |
+
)
|
| 402 |
+
|
| 403 |
+
input_features = input_features.to(dtype=self.conv1.weight.dtype, device=self.conv1.weight.device)
|
| 404 |
+
inputs_embeds = nn.functional.gelu(self.conv1(input_features))
|
| 405 |
+
inputs_embeds = nn.functional.gelu(self.conv2(inputs_embeds))
|
| 406 |
+
inputs_embeds = inputs_embeds.permute(0, 2, 1)
|
| 407 |
+
|
| 408 |
+
# Apply initial FDDT if configured
|
| 409 |
+
if (self.config.use_dicow_encoder and
|
| 410 |
+
self.config.use_fddt and
|
| 411 |
+
self.config.use_pre_pos_fddt and
|
| 412 |
+
hasattr(self, 'initial_fddt')):
|
| 413 |
+
inputs_embeds = self.initial_fddt(inputs_embeds, stno_mask)
|
| 414 |
+
|
| 415 |
+
embed_pos = self.embed_positions.weight
|
| 416 |
+
hidden_states = (inputs_embeds + embed_pos).to(inputs_embeds.dtype)
|
| 417 |
+
hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
|
| 418 |
+
|
| 419 |
+
for idx, encoder_layer in enumerate(self.layers):
|
| 420 |
+
|
| 421 |
+
if (self.config.use_dicow_encoder and
|
| 422 |
+
self.config.use_fddt and
|
| 423 |
+
hasattr(self, 'fddts') and
|
| 424 |
+
idx < len(self.fddts)):
|
| 425 |
+
hidden_states = self.fddts[idx](hidden_states, stno_mask)
|
| 426 |
+
|
| 427 |
+
layer_outputs = encoder_layer(
|
| 428 |
+
hidden_states,
|
| 429 |
+
attention_mask=attention_mask,
|
| 430 |
+
layer_head_mask=None,
|
| 431 |
+
)
|
| 432 |
+
hidden_states = layer_outputs[0]
|
| 433 |
+
|
| 434 |
+
hidden_states = self.layer_norm(hidden_states)
|
| 435 |
+
|
| 436 |
+
return BaseModelOutput(
|
| 437 |
+
last_hidden_state=hidden_states,
|
| 438 |
+
)
|
| 439 |
+
|
| 440 |
+
|
| 441 |
+
# Ignore copy
|
| 442 |
+
def _get_feat_extract_output_lengths(self, input_lengths: torch.LongTensor):
|
| 443 |
+
"""
|
| 444 |
+
Computes the output length of the convolutional layers and the output length of the audio encoder
|
| 445 |
+
"""
|
| 446 |
+
input_lengths = (input_lengths - 1) // 2 + 1
|
| 447 |
+
output_lengths = (input_lengths - 2) // 2 + 1
|
| 448 |
+
return input_lengths, output_lengths
|
| 449 |
+
|
| 450 |
+
|
| 451 |
+
class VoxtralMultiModalProjector(nn.Module):
|
| 452 |
+
def __init__(self, config: VoxtralConfig):
|
| 453 |
+
super().__init__()
|
| 454 |
+
self.linear_1 = nn.Linear(config.audio_config.intermediate_size, config.text_config.hidden_size, bias=False)
|
| 455 |
+
self.act = ACT2FN[config.projector_hidden_act]
|
| 456 |
+
self.linear_2 = nn.Linear(config.text_config.hidden_size, config.text_config.hidden_size, bias=False)
|
| 457 |
+
|
| 458 |
+
def forward(self, audio_features):
|
| 459 |
+
hidden_states = self.linear_1(audio_features)
|
| 460 |
+
hidden_states = self.act(hidden_states)
|
| 461 |
+
hidden_states = self.linear_2(hidden_states)
|
| 462 |
+
return hidden_states
|
| 463 |
+
|
| 464 |
+
|
| 465 |
+
@auto_docstring(
|
| 466 |
+
custom_intro="""
|
| 467 |
+
The Voxtral model, which consists of Whisper encoder, a multi-modal projector and a LLama language model.
|
| 468 |
+
"""
|
| 469 |
+
)
|
| 470 |
+
class DixtralForConditionalGeneration(DixtralPreTrainedModel, GenerationMixin):
|
| 471 |
+
_tied_weights_keys = ["lm_head.weight"]
|
| 472 |
+
_tp_plan = {"lm_head": "colwise_rep"}
|
| 473 |
+
_pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
|
| 474 |
+
_keep_in_fp32_modules_strict = ["embed_positions"]
|
| 475 |
+
|
| 476 |
+
def __init__(self, config):
|
| 477 |
+
super().__init__(config)
|
| 478 |
+
self.vocab_size = config.text_config.vocab_size
|
| 479 |
+
self.audio_tower = DixtralEncoder(config.audio_config)
|
| 480 |
+
self.language_model = AutoModelForCausalLM.from_config(config.text_config)
|
| 481 |
+
self.multi_modal_projector = VoxtralMultiModalProjector(config)
|
| 482 |
+
|
| 483 |
+
self.num_soft_prompts = config.num_soft_prompts
|
| 484 |
+
if self.num_soft_prompts > 0:
|
| 485 |
+
self.soft_prompt_token_id = getattr(config, "soft_prompt_token_id", 23)
|
| 486 |
+
|
| 487 |
+
self.soft_prompt = nn.Parameter(
|
| 488 |
+
torch.randn(1, self.num_soft_prompts, config.text_config.hidden_size)
|
| 489 |
+
)
|
| 490 |
+
|
| 491 |
+
self._init_dicow_components(config)
|
| 492 |
+
# Initialize weights and apply final processing
|
| 493 |
+
self.post_init()
|
| 494 |
+
|
| 495 |
+
def _init_dicow_components(self, config):
|
| 496 |
+
self.ctc_weight = config.audio_config.ctc_weight
|
| 497 |
+
|
| 498 |
+
# Additional layers for CTC
|
| 499 |
+
if config.audio_config.additional_layer and self.ctc_weight > 0.0:
|
| 500 |
+
custom_conf = copy.deepcopy(config.audio_config)
|
| 501 |
+
custom_conf.d_model = config.text_config.hidden_size
|
| 502 |
+
custom_conf.encoder_attention_heads = config.text_config.num_attention_heads
|
| 503 |
+
custom_conf.encoder_ffn_dim = custom_conf.d_model * 2
|
| 504 |
+
self.additional_layer = VoxtralEncoderLayer(custom_conf)
|
| 505 |
+
|
| 506 |
+
if config.audio_config.additional_self_attention_layer and self.ctc_weight > 0.0:
|
| 507 |
+
self.additional_self_attention_layer = VoxtralAttention(
|
| 508 |
+
embed_dim=config.text_config.hidden_size,
|
| 509 |
+
num_heads=config.text_config.num_attention_heads,
|
| 510 |
+
dropout=config.text_config.attention_dropout,
|
| 511 |
+
config=config.audio_config, # Fixed: pass audio_config which is VoxtralConfig
|
| 512 |
+
)
|
| 513 |
+
|
| 514 |
+
# CTC head
|
| 515 |
+
if self.ctc_weight > 0.0:
|
| 516 |
+
self.ctc_lm_head = nn.Linear(config.text_config.hidden_size, config.text_config.vocab_size, bias=False)
|
| 517 |
+
self.ctc_lm_head.weight = self.language_model.get_input_embeddings().weight
|
| 518 |
+
def get_input_embeddings(self):
|
| 519 |
+
return self.language_model.get_input_embeddings()
|
| 520 |
+
|
| 521 |
+
def set_input_embeddings(self, value):
|
| 522 |
+
self.language_model.set_input_embeddings(value)
|
| 523 |
+
|
| 524 |
+
def get_output_embeddings(self):
|
| 525 |
+
return self.language_model.get_output_embeddings()
|
| 526 |
+
|
| 527 |
+
def set_output_embeddings(self, new_embeddings):
|
| 528 |
+
self.language_model.set_output_embeddings(new_embeddings)
|
| 529 |
+
|
| 530 |
+
def set_decoder(self, decoder):
|
| 531 |
+
self.language_model.set_decoder(decoder)
|
| 532 |
+
|
| 533 |
+
def get_decoder(self):
|
| 534 |
+
return self.language_model.get_decoder()
|
| 535 |
+
|
| 536 |
+
def get_audio_embeds(self, input_features: torch.FloatTensor, stno_mask: torch.FloatTensor):
|
| 537 |
+
"""
|
| 538 |
+
This method is used to get the audio embeddings from input features (a log mel spectrogram), meaning inferring the audio encoder and the multi-modal projector.
|
| 539 |
+
Args:
|
| 540 |
+
input_features (`torch.FloatTensor`):
|
| 541 |
+
Float values of mel features extracted from the raw speech waveform. Raw speech waveform can be
|
| 542 |
+
obtained by loading a `.flac` or `.wav` audio file into an array of type `list[float]` or a
|
| 543 |
+
`numpy.ndarray`, *e.g.* via the soundfile library (`pip install soundfile`). To prepare the array into
|
| 544 |
+
`input_features`, the [`AutoFeatureExtractor`] should be used for extracting the mel features, padding
|
| 545 |
+
and conversion into a tensor of type `torch.FloatTensor`. See [`~WhisperFeatureExtractor.__call__`]
|
| 546 |
+
|
| 547 |
+
Returns:
|
| 548 |
+
`torch.FloatTensor`:
|
| 549 |
+
The audio embeddings.
|
| 550 |
+
"""
|
| 551 |
+
audio_outputs = self.audio_tower(input_features, stno_mask=stno_mask)
|
| 552 |
+
audio_hidden_states = audio_outputs.last_hidden_state
|
| 553 |
+
audio_hidden_states = audio_hidden_states.reshape(-1, self.config.audio_config.intermediate_size)
|
| 554 |
+
audio_embeds = self.multi_modal_projector(audio_hidden_states)
|
| 555 |
+
return audio_embeds
|
| 556 |
+
|
| 557 |
+
def set_tokenizer(self, tokenizer):
|
| 558 |
+
self.tokenizer = tokenizer
|
| 559 |
+
|
| 560 |
+
|
| 561 |
+
def possibly_update_last_hidden_states(self, hidden_states):
|
| 562 |
+
"""DiCoW post-processing for CTC"""
|
| 563 |
+
if not self.config.audio_config.use_dicow_encoder:
|
| 564 |
+
return hidden_states
|
| 565 |
+
|
| 566 |
+
if hasattr(self, "additional_layer"):
|
| 567 |
+
hidden_states, _ = self.additional_layer(
|
| 568 |
+
hidden_states,
|
| 569 |
+
attention_mask=None,
|
| 570 |
+
layer_head_mask=None,
|
| 571 |
+
output_attentions=False,
|
| 572 |
+
)
|
| 573 |
+
elif hasattr(self, "additional_self_attention_layer"):
|
| 574 |
+
hidden_states, _ = self.additional_self_attention_layer(
|
| 575 |
+
hidden_states,
|
| 576 |
+
attention_mask=None,
|
| 577 |
+
layer_head_mask=None,
|
| 578 |
+
output_attentions=False,
|
| 579 |
+
)
|
| 580 |
+
|
| 581 |
+
return hidden_states
|
| 582 |
+
|
| 583 |
+
def get_enc_logits(self, hidden_states):
|
| 584 |
+
"""
|
| 585 |
+
Get CTC logits from encoder hidden states.
|
| 586 |
+
Applies optional additional processing layer and projects to vocabulary.
|
| 587 |
+
|
| 588 |
+
Args:
|
| 589 |
+
hidden_states: Encoder output hidden states
|
| 590 |
+
|
| 591 |
+
Returns:
|
| 592 |
+
logits: CTC logits of shape (batch_size, seq_len, vocab_size + 1)
|
| 593 |
+
"""
|
| 594 |
+
hidden_states = self.possibly_update_last_hidden_states(hidden_states)
|
| 595 |
+
logits = self.ctc_lm_head(hidden_states)
|
| 596 |
+
return logits
|
| 597 |
+
|
| 598 |
+
def right_pad_labels(self, labels, pad_value=-100):
|
| 599 |
+
"""
|
| 600 |
+
labels: (B, L) tensor possibly left/right padded
|
| 601 |
+
returns: right-padded labels only
|
| 602 |
+
"""
|
| 603 |
+
B, L = labels.shape
|
| 604 |
+
new_labels = torch.full_like(labels, pad_value)
|
| 605 |
+
max_len = 1
|
| 606 |
+
for b in range(B):
|
| 607 |
+
valid = labels[b][labels[b] != pad_value]
|
| 608 |
+
max_len = max(max_len, len(valid))
|
| 609 |
+
new_labels[b, :valid.numel()] = valid
|
| 610 |
+
|
| 611 |
+
new_labels = new_labels[:, :max_len]
|
| 612 |
+
|
| 613 |
+
return new_labels
|
| 614 |
+
|
| 615 |
+
def get_ctc_loss(self, logits, labels, input_lengths):
|
| 616 |
+
|
| 617 |
+
"""Compute CTC loss for DiCoW"""
|
| 618 |
+
if labels.max() >= self.config.text_config.vocab_size:
|
| 619 |
+
raise ValueError(f"Label values must be <= vocab_size: {self.config.text_config.vocab_size}")
|
| 620 |
+
|
| 621 |
+
# Assuming that padded tokens are filled with -100
|
| 622 |
+
labels_mask = labels >= 0
|
| 623 |
+
target_lengths = labels_mask.sum(-1)
|
| 624 |
+
|
| 625 |
+
# CTC loss doesn't support fp16
|
| 626 |
+
log_probs = nn.functional.log_softmax(logits, dim=-1, dtype=torch.float32).transpose(0, 1)
|
| 627 |
+
|
| 628 |
+
with torch.backends.cudnn.flags(enabled=True):
|
| 629 |
+
ctc_loss = nn.functional.ctc_loss(
|
| 630 |
+
log_probs,
|
| 631 |
+
labels,
|
| 632 |
+
input_lengths,
|
| 633 |
+
target_lengths,
|
| 634 |
+
blank=logits.shape[-1] - 1,
|
| 635 |
+
reduction=self.config.audio_config.ctc_loss_reduction,
|
| 636 |
+
zero_infinity=True,
|
| 637 |
+
)
|
| 638 |
+
|
| 639 |
+
return ctc_loss
|
| 640 |
+
|
| 641 |
+
@can_return_tuple
|
| 642 |
+
@auto_docstring
|
| 643 |
+
def forward(
|
| 644 |
+
self,
|
| 645 |
+
input_ids: Optional[torch.LongTensor] = None,
|
| 646 |
+
input_features: Optional[torch.FloatTensor] = None,
|
| 647 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 648 |
+
position_ids: Optional[torch.LongTensor] = None,
|
| 649 |
+
past_key_values: Optional[Cache] = None,
|
| 650 |
+
inputs_embeds: Optional[torch.FloatTensor] = None,
|
| 651 |
+
labels: Optional[torch.LongTensor] = None,
|
| 652 |
+
use_cache: Optional[bool] = None,
|
| 653 |
+
cache_position: Optional[torch.LongTensor] = None,
|
| 654 |
+
logits_to_keep: Union[int, torch.Tensor] = 0,
|
| 655 |
+
stno_mask=None,
|
| 656 |
+
**kwargs: Unpack[TransformersKwargs],
|
| 657 |
+
) -> CausalLMOutputWithPast:
|
| 658 |
+
r"""
|
| 659 |
+
Example:
|
| 660 |
+
|
| 661 |
+
```python
|
| 662 |
+
>>> from transformers import VoxtralForConditionalGeneration, AutoProcessor
|
| 663 |
+
>>> import torch
|
| 664 |
+
|
| 665 |
+
>>> device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 666 |
+
>>> repo_id = "mistralai/Voxtral-Mini-3B-2507"
|
| 667 |
+
|
| 668 |
+
>>> processor = AutoProcessor.from_pretrained(repo_id)
|
| 669 |
+
>>> model = VoxtralForConditionalGeneration.from_pretrained(repo_id, torch_dtype=torch.bfloat16, device_map=device)
|
| 670 |
+
|
| 671 |
+
>>> conversation = [
|
| 672 |
+
{
|
| 673 |
+
"role": "user",
|
| 674 |
+
"content": [
|
| 675 |
+
{
|
| 676 |
+
"type": "audio",
|
| 677 |
+
"url": "https://huggingface.co/datasets/hf-internal-testing/dummy-audio-samples/resolve/main/dude_where_is_my_car.wav",
|
| 678 |
+
},
|
| 679 |
+
{"type": "text", "text": "What can you tell me about this audio?"},
|
| 680 |
+
],
|
| 681 |
+
}
|
| 682 |
+
]
|
| 683 |
+
|
| 684 |
+
>>> inputs = processor.apply_chat_template(conversation)
|
| 685 |
+
>>> inputs = inputs.to(device, dtype=torch.bfloat16)
|
| 686 |
+
|
| 687 |
+
>>> outputs = model.generate(**inputs, max_new_tokens=30)
|
| 688 |
+
>>> processor.batch_decode(outputs[:, inputs.input_ids.shape[1]:], skip_special_tokens=True)
|
| 689 |
+
["This audio is a humorous conversation between two friends, likely in English, where one of them is trying to figure out what the other's tattoo says."]
|
| 690 |
+
```"""
|
| 691 |
+
if inputs_embeds is None:
|
| 692 |
+
inputs_embeds = self.get_input_embeddings()(input_ids)
|
| 693 |
+
|
| 694 |
+
ctc_loss = None
|
| 695 |
+
if input_features is not None:
|
| 696 |
+
# Get audio encoder outputs
|
| 697 |
+
audio_outputs = self.audio_tower(input_features, stno_mask=stno_mask)
|
| 698 |
+
audio_hidden_states = audio_outputs.last_hidden_state
|
| 699 |
+
|
| 700 |
+
# Project audio features for language model
|
| 701 |
+
audio_hidden_states_flat = audio_hidden_states.reshape(-1, self.config.audio_config.intermediate_size)
|
| 702 |
+
audio_embeds_flat = self.multi_modal_projector(audio_hidden_states_flat)
|
| 703 |
+
|
| 704 |
+
# Replace text-audio token placeholders with audio embeddings
|
| 705 |
+
audio_token_mask = input_ids == self.config.audio_token_id
|
| 706 |
+
inputs_embeds[audio_token_mask] = audio_embeds_flat
|
| 707 |
+
|
| 708 |
+
if self.num_soft_prompts > 0:
|
| 709 |
+
prompt_mask = (input_ids == self.soft_prompt_token_id)
|
| 710 |
+
|
| 711 |
+
if prompt_mask.any():
|
| 712 |
+
batch_size = inputs_embeds.shape[0]
|
| 713 |
+
|
| 714 |
+
# Expand the learned soft prompts to [Batch_Size, Num_Soft_Tokens, Hidden_Size]
|
| 715 |
+
# Then flatten to [Batch_Size * Num_Soft_Tokens, Hidden_Size] to match the mask
|
| 716 |
+
prompts_expanded = self.soft_prompt.expand(batch_size, -1, -1).reshape(-1,
|
| 717 |
+
self.config.text_config.hidden_size)
|
| 718 |
+
|
| 719 |
+
# Replace embeddings
|
| 720 |
+
inputs_embeds[prompt_mask] = prompts_expanded
|
| 721 |
+
|
| 722 |
+
# Compute CTC loss on projected embeddings if configured
|
| 723 |
+
if (self.config.audio_config.use_dicow_encoder and
|
| 724 |
+
self.config.audio_config.ctc_weight > 0.0 and
|
| 725 |
+
labels is not None and
|
| 726 |
+
self.training and
|
| 727 |
+
audio_token_mask is not None) or hasattr(self, "ctc_rescorer"):
|
| 728 |
+
|
| 729 |
+
# Create tensor with shape of input_ids filled with zeros
|
| 730 |
+
batch_size, seq_len = input_ids.shape
|
| 731 |
+
hidden_dim = audio_embeds_flat.shape[-1]
|
| 732 |
+
ctc_embeds = torch.empty(
|
| 733 |
+
batch_size, seq_len, hidden_dim,
|
| 734 |
+
device=audio_embeds_flat.device,
|
| 735 |
+
dtype=audio_embeds_flat.dtype
|
| 736 |
+
)
|
| 737 |
+
|
| 738 |
+
# Fill with audio_embeds at audio_token positions
|
| 739 |
+
ctc_embeds[audio_token_mask] = audio_embeds_flat
|
| 740 |
+
ctc_embeds_detached = ctc_embeds.detach()
|
| 741 |
+
|
| 742 |
+
# 2. Force it to require gradients so the additional_layer
|
| 743 |
+
# builds a backward graph for its own weights
|
| 744 |
+
ctc_embeds_detached.requires_grad_(True)
|
| 745 |
+
|
| 746 |
+
# Remove values outside maximum valid range using audio_mask
|
| 747 |
+
enc_output_lens = audio_token_mask.sum(dim=1)
|
| 748 |
+
max_valid_len = enc_output_lens.max().item()
|
| 749 |
+
first_audio_token = audio_token_mask.int().argmax(dim=1).min().item() # First True position per batch
|
| 750 |
+
ctc_embeds = ctc_embeds[:, first_audio_token:first_audio_token+max_valid_len, :]
|
| 751 |
+
|
| 752 |
+
# Get encoder logits for CTC
|
| 753 |
+
enc_logits = self.get_enc_logits(ctc_embeds)
|
| 754 |
+
|
| 755 |
+
if hasattr(self, "ctc_rescorer"):
|
| 756 |
+
rescorer = CTCRescorerLogitsProcessorWithPruning(
|
| 757 |
+
enc_logits,
|
| 758 |
+
torch.full((enc_logits.shape[0],), fill_value=enc_logits.shape[1],
|
| 759 |
+
device=enc_logits.device),
|
| 760 |
+
enc_logits.shape[-1] - 1,
|
| 761 |
+
self.generation_config.pad_token_id,
|
| 762 |
+
self.generation_config.eos_token_id,
|
| 763 |
+
self.generation_config.bos_token_id,
|
| 764 |
+
self.tokenizer,
|
| 765 |
+
0,
|
| 766 |
+
self.generation_config.ctc_weight,
|
| 767 |
+
self.generation_config.num_beams,
|
| 768 |
+
False,
|
| 769 |
+
)
|
| 770 |
+
self.ctc_rescorer.set_func(func=rescorer)
|
| 771 |
+
|
| 772 |
+
if labels is not None:
|
| 773 |
+
# Prepare encoder labels
|
| 774 |
+
enc_labels = labels.clone()
|
| 775 |
+
|
| 776 |
+
# Replace EOS tokens with ignore index
|
| 777 |
+
enc_labels[enc_labels == self.config.text_config.eos_token_id] = -100
|
| 778 |
+
enc_labels = self.right_pad_labels(enc_labels)
|
| 779 |
+
|
| 780 |
+
# Compute CTC loss
|
| 781 |
+
ctc_loss = self.get_ctc_loss(enc_logits, enc_labels, enc_output_lens)
|
| 782 |
+
|
| 783 |
+
outputs: BaseModelOutputWithPast = self.language_model(
|
| 784 |
+
attention_mask=attention_mask,
|
| 785 |
+
position_ids=position_ids,
|
| 786 |
+
past_key_values=past_key_values,
|
| 787 |
+
inputs_embeds=inputs_embeds,
|
| 788 |
+
labels=labels,
|
| 789 |
+
use_cache=use_cache,
|
| 790 |
+
cache_position=cache_position,
|
| 791 |
+
logits_to_keep=logits_to_keep,
|
| 792 |
+
**kwargs,
|
| 793 |
+
)
|
| 794 |
+
|
| 795 |
+
if ctc_loss is not None and outputs.loss is not None:
|
| 796 |
+
if wandb.run is not None:
|
| 797 |
+
wandb.log({"dec_loss": outputs.loss, "ctc_loss": ctc_loss})
|
| 798 |
+
total_loss = outputs.loss + self.config.audio_config.ctc_weight * ctc_loss
|
| 799 |
+
outputs.loss = total_loss
|
| 800 |
+
elif ctc_loss is not None:
|
| 801 |
+
outputs.loss = self.config.audio_config.ctc_weight * ctc_loss
|
| 802 |
+
|
| 803 |
+
|
| 804 |
+
return outputs
|
| 805 |
+
|
| 806 |
+
def prepare_inputs_for_generation(self, *args, **kwargs):
|
| 807 |
+
# Overwritten -- we should not pass input_features/stno_mask when in cached decoding stage
|
| 808 |
+
|
| 809 |
+
input_features = kwargs.pop("input_features", None)
|
| 810 |
+
stno_mask = kwargs.pop("stno_mask", None)
|
| 811 |
+
cache_position = kwargs.get("cache_position")
|
| 812 |
+
|
| 813 |
+
model_inputs = super().prepare_inputs_for_generation(*args, **kwargs)
|
| 814 |
+
|
| 815 |
+
if cache_position is not None and cache_position[0] == 0:
|
| 816 |
+
# Only pass audio inputs on the first (prefill) step
|
| 817 |
+
model_inputs["input_features"] = input_features
|
| 818 |
+
model_inputs["stno_mask"] = stno_mask
|
| 819 |
+
|
| 820 |
+
return model_inputs
|
| 821 |
+
|
| 822 |
+
|
| 823 |
+
def _get_logits_processor(
|
| 824 |
+
self,
|
| 825 |
+
generation_config: GenerationConfig,
|
| 826 |
+
input_ids_seq_length: Optional[int] = None,
|
| 827 |
+
encoder_input_ids: torch.LongTensor = None,
|
| 828 |
+
prefix_allowed_tokens_fn: Optional[Callable[[int, torch.Tensor], list[int]]] = None,
|
| 829 |
+
logits_processor: Optional[LogitsProcessorList] = None,
|
| 830 |
+
device: Optional[str] = None,
|
| 831 |
+
model_kwargs: Optional[dict[str, Any]] = None,
|
| 832 |
+
negative_prompt_ids: Optional[torch.Tensor] = None,
|
| 833 |
+
negative_prompt_attention_mask: Optional[torch.Tensor] = None,
|
| 834 |
+
) -> LogitsProcessorList:
|
| 835 |
+
# pylint: disable=no-member
|
| 836 |
+
gen_config_copy = copy.deepcopy(generation_config)
|
| 837 |
+
processors = super()._get_logits_processor(
|
| 838 |
+
gen_config_copy,
|
| 839 |
+
input_ids_seq_length,
|
| 840 |
+
encoder_input_ids,
|
| 841 |
+
prefix_allowed_tokens_fn,
|
| 842 |
+
logits_processor,
|
| 843 |
+
device,
|
| 844 |
+
model_kwargs,
|
| 845 |
+
negative_prompt_ids,
|
| 846 |
+
negative_prompt_attention_mask,
|
| 847 |
+
)
|
| 848 |
+
if hasattr(generation_config, "ctc_weight") and generation_config.ctc_weight > 0:
|
| 849 |
+
self.ctc_rescorer = CTCProcessorDummy
|
| 850 |
+
processors.append(self.ctc_rescorer)
|
| 851 |
+
return processors
|
| 852 |
+
|
| 853 |
+
@torch.no_grad()
|
| 854 |
+
def decode_ctc(
|
| 855 |
+
self,
|
| 856 |
+
input_ids: torch.LongTensor,
|
| 857 |
+
input_features: torch.FloatTensor,
|
| 858 |
+
stno_mask: Optional[torch.Tensor] = None,
|
| 859 |
+
) -> tuple[None, torch.LongTensor]:
|
| 860 |
+
"""
|
| 861 |
+
Performs greedy CTC decoding on the audio input.
|
| 862 |
+
"""
|
| 863 |
+
|
| 864 |
+
audio_outputs = self.audio_tower(input_features, stno_mask=stno_mask)
|
| 865 |
+
audio_hidden_states = audio_outputs.last_hidden_state
|
| 866 |
+
|
| 867 |
+
# Project audio features for language model
|
| 868 |
+
audio_hidden_states_flat = audio_hidden_states.reshape(-1, self.config.audio_config.intermediate_size)
|
| 869 |
+
audio_embeds_flat = self.multi_modal_projector(audio_hidden_states_flat)
|
| 870 |
+
|
| 871 |
+
# Replace text-audio token placeholders with audio embeddings
|
| 872 |
+
audio_token_mask = input_ids == self.config.audio_token_id
|
| 873 |
+
|
| 874 |
+
# Create tensor with shape of input_ids filled with zeros
|
| 875 |
+
batch_size, seq_len = input_ids.shape
|
| 876 |
+
hidden_dim = audio_embeds_flat.shape[-1]
|
| 877 |
+
ctc_embeds = torch.empty(
|
| 878 |
+
batch_size, seq_len, hidden_dim,
|
| 879 |
+
device=audio_embeds_flat.device,
|
| 880 |
+
dtype=audio_embeds_flat.dtype
|
| 881 |
+
)
|
| 882 |
+
|
| 883 |
+
# Fill with audio_embeds at audio_token positions
|
| 884 |
+
ctc_embeds[audio_token_mask] = audio_embeds_flat
|
| 885 |
+
|
| 886 |
+
# Remove values outside maximum valid range using audio_mask
|
| 887 |
+
enc_output_lens = audio_token_mask.sum(dim=1)
|
| 888 |
+
max_valid_len = enc_output_lens.max().item()
|
| 889 |
+
first_audio_token = audio_token_mask.int().argmax(dim=1).min().item() # First True position per batch
|
| 890 |
+
ctc_embeds = ctc_embeds[:, first_audio_token:first_audio_token + max_valid_len, :]
|
| 891 |
+
|
| 892 |
+
# Get encoder logits for CTC
|
| 893 |
+
logits = self.get_enc_logits(ctc_embeds)
|
| 894 |
+
|
| 895 |
+
# 4. Greedy Decoding
|
| 896 |
+
predicted_ids = torch.argmax(logits, dim=-1)
|
| 897 |
+
|
| 898 |
+
# Blank token is the last index in the vocabulary (vocab_size - 1)
|
| 899 |
+
# Based on: blank=logits.shape[-1] - 1 in get_ctc_loss
|
| 900 |
+
blank_id = self.config.text_config.vocab_size - 1
|
| 901 |
+
|
| 902 |
+
sequences = []
|
| 903 |
+
|
| 904 |
+
for batch_idx in range(batch_size):
|
| 905 |
+
ids = predicted_ids[batch_idx].cpu().tolist()
|
| 906 |
+
|
| 907 |
+
# CTC Collapse:
|
| 908 |
+
# 1. Merge adjacent duplicates
|
| 909 |
+
# 2. Remove blank tokens
|
| 910 |
+
collapsed_ids = []
|
| 911 |
+
prev_id = -1
|
| 912 |
+
|
| 913 |
+
for token_id in ids:
|
| 914 |
+
if token_id != prev_id:
|
| 915 |
+
if token_id != blank_id:
|
| 916 |
+
collapsed_ids.append(token_id)
|
| 917 |
+
prev_id = token_id
|
| 918 |
+
|
| 919 |
+
sequences.append(torch.tensor(collapsed_ids, dtype=torch.long))
|
| 920 |
+
|
| 921 |
+
return None, torch.nn.utils.rnn.pad_sequence(sequences, batch_first=True, padding_value=-100).to(input_ids.device)
|
| 922 |
+
|
| 923 |
+
__all__ = ["DixtralPreTrainedModel", "DixtralEncoder", "DixtralForConditionalGeneration"]
|