{ "model_name_or_path": "./train/qwen_freeze_emb", "text_vocab_size": 151665, "s_id": 151665, "t_id": 151666, "hubert_offset": 151667, "total_vocab_size": 152167, "codebook_size": 500, "patch_mode": "static", "patch_size": 8, "max_patch_length": null, "pooling": "mean", "local_layers": 1, "local_heads": 8, "cross_heads": 8, "bidirectional_local_encoder": false, "input_embeddings_frozen": true, "freeze_lm_head": false, "objective": "speech_patches_to_text_only", "patch_encoder_state": "speech_patch_encoder.pt" }