Training in progress, step 500
Browse files- config.json +10 -4
- generation_config.json +1 -1
- model.safetensors +2 -2
- training_args.bin +2 -2
config.json
CHANGED
|
@@ -64,25 +64,31 @@
|
|
| 64 |
"type": "audio"
|
| 65 |
}
|
| 66 |
},
|
|
|
|
| 67 |
"dtype": "bfloat16",
|
| 68 |
"encoder_dim": 1280,
|
|
|
|
| 69 |
"inference_diversity_penalty": 0.0,
|
| 70 |
"inference_warmup_tokens": 10,
|
| 71 |
"label_smoothing": 0.0,
|
| 72 |
"llm_dim": 2048,
|
| 73 |
-
"max_new_tokens":
|
| 74 |
-
"min_new_tokens":
|
| 75 |
"model_dtype": "bfloat16",
|
| 76 |
"model_type": "asr_model",
|
|
|
|
|
|
|
| 77 |
"pipeline_tag": "automatic-speech-recognition",
|
| 78 |
-
"projector_dropout": 0.
|
| 79 |
"projector_hidden_dim": null,
|
| 80 |
"projector_init_std": 0.02,
|
| 81 |
"projector_input_noise": 0.02,
|
| 82 |
"projector_num_layers": 2,
|
| 83 |
"projector_pool_stride": 2,
|
| 84 |
"projector_type": "swiglu",
|
|
|
|
| 85 |
"system_prompt": "/no_think /system_override",
|
|
|
|
| 86 |
"text_config": {
|
| 87 |
"_name_or_path": "HuggingFaceTB/SmolLM3-3B",
|
| 88 |
"architectures": [
|
|
@@ -191,7 +197,7 @@
|
|
| 191 |
"vocab_size": 128257
|
| 192 |
},
|
| 193 |
"text_model_id": "HuggingFaceTB/SmolLM3-3B",
|
| 194 |
-
"transformers_version": "4.57.
|
| 195 |
"use_cache": false,
|
| 196 |
"use_specaugment": true,
|
| 197 |
"user_prompt": "Transcribe: <audio>",
|
|
|
|
| 64 |
"type": "audio"
|
| 65 |
}
|
| 66 |
},
|
| 67 |
+
"decoder_model_name": "HuggingFaceTB/SmolLM3-3B",
|
| 68 |
"dtype": "bfloat16",
|
| 69 |
"encoder_dim": 1280,
|
| 70 |
+
"encoder_model_name": "openai/whisper-large-v3-turbo",
|
| 71 |
"inference_diversity_penalty": 0.0,
|
| 72 |
"inference_warmup_tokens": 10,
|
| 73 |
"label_smoothing": 0.0,
|
| 74 |
"llm_dim": 2048,
|
| 75 |
+
"max_new_tokens": 256,
|
| 76 |
+
"min_new_tokens": 0,
|
| 77 |
"model_dtype": "bfloat16",
|
| 78 |
"model_type": "asr_model",
|
| 79 |
+
"num_experts": 4,
|
| 80 |
+
"num_experts_per_tok": 2,
|
| 81 |
"pipeline_tag": "automatic-speech-recognition",
|
| 82 |
+
"projector_dropout": 0.1,
|
| 83 |
"projector_hidden_dim": null,
|
| 84 |
"projector_init_std": 0.02,
|
| 85 |
"projector_input_noise": 0.02,
|
| 86 |
"projector_num_layers": 2,
|
| 87 |
"projector_pool_stride": 2,
|
| 88 |
"projector_type": "swiglu",
|
| 89 |
+
"router_aux_loss_coef": 0.01,
|
| 90 |
"system_prompt": "/no_think /system_override",
|
| 91 |
+
"temperature": 0.1,
|
| 92 |
"text_config": {
|
| 93 |
"_name_or_path": "HuggingFaceTB/SmolLM3-3B",
|
| 94 |
"architectures": [
|
|
|
|
| 197 |
"vocab_size": 128257
|
| 198 |
},
|
| 199 |
"text_model_id": "HuggingFaceTB/SmolLM3-3B",
|
| 200 |
+
"transformers_version": "4.57.3",
|
| 201 |
"use_cache": false,
|
| 202 |
"use_specaugment": true,
|
| 203 |
"user_prompt": "Transcribe: <audio>",
|
generation_config.json
CHANGED
|
@@ -4,5 +4,5 @@
|
|
| 4 |
"pad_token_id": 128004,
|
| 5 |
"temperature": 0.6,
|
| 6 |
"top_p": 0.95,
|
| 7 |
-
"transformers_version": "4.57.
|
| 8 |
}
|
|
|
|
| 4 |
"pad_token_id": 128004,
|
| 5 |
"temperature": 0.6,
|
| 6 |
"top_p": 0.95,
|
| 7 |
+
"transformers_version": "4.57.3"
|
| 8 |
}
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9b12132b4625446fcbd9b3d942aebec4cf5686e57c8683526bb1e7e5564f5084
|
| 3 |
+
size 76022400
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:161661b994f6ca586957ca389329090fcf4e91738ad2b4c103e427316d33abcd
|
| 3 |
+
size 5905
|