Upload train_config.json with huggingface_hub

train_config.json

@@ -0,0 +1,61 @@
+{
+    "output_dir": "exp/omnivoice_wordcontrol",
+    "data_config": "configs/data_config_wordcontrol.json",
+    "llm_name_or_path": "Qwen/Qwen3-0.6B",
+    "audio_vocab_size": 1025,
+    "audio_mask_id": 1024,
+    "num_audio_codebook": 8,
+    "audio_codebook_weights": [
+        8,
+        8,
+        6,
+        6,
+        4,
+        4,
+        2,
+        2
+    ],
+    "drop_cond_ratio": 0.1,
+    "prompt_ratio_range": [
+        0.0,
+        0.3
+    ],
+    "mask_ratio_range": [
+        0.0,
+        1.0
+    ],
+    "language_ratio": 0.8,
+    "use_pinyin_ratio": 0.0,
+    "instruct_ratio": 0.0,
+    "only_instruct_ratio": 0.0,
+    "word_control": true,
+    "wc_drop_all_ratio": 0.15,
+    "wc_word_drop_ratio": 0.2,
+    "wc_attr_keep_ratio": 0.7,
+    "resume_from_checkpoint": null,
+    "init_from_checkpoint": "k2-fsa/OmniVoice",
+    "learning_rate": 5e-05,
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "steps": 40000,
+    "seed": 42,
+    "lr_scheduler_type": "cosine",
+    "warmup_type": "ratio",
+    "warmup_ratio": 0.02,
+    "warmup_steps": 2000,
+    "batch_tokens": 4096,
+    "gradient_accumulation_steps": 2,
+    "num_workers": 4,
+    "mixed_precision": "bf16",
+    "allow_tf32": true,
+    "use_deepspeed": false,
+    "deepspeed_config": null,
+    "attn_implementation": "flex_attention",
+    "max_sample_tokens": 2000,
+    "min_sample_tokens": 50,
+    "max_batch_size": 64,
+    "logging_steps": 50,
+    "eval_steps": 2000,
+    "save_steps": 2000,
+    "keep_last_n_checkpoints": 3
+}