| { |
| "architectures": [ |
| "JaisASRForConditionalGeneration" |
| ], |
| "audio_config": { |
| "activation_dropout": 0, |
| "activation_function": "gelu", |
| "attention_dropout": 0, |
| "conv_chunksize": 500, |
| "d_model": 1280, |
| "downsample_hidden_size": 480, |
| "dropout": 0, |
| "dtype": "bfloat16", |
| "encoder_attention_heads": 20, |
| "encoder_ffn_dim": 5120, |
| "encoder_layers": 32, |
| "initializer_range": 0.02, |
| "max_source_positions": 1500, |
| "model_type": "qwen3_omni_moe_audio_encoder", |
| "n_window": 50, |
| "n_window_infer": 800, |
| "num_hidden_layers": 32, |
| "num_mel_bins": 128, |
| "output_dim": 2048, |
| "scale_embedding": false, |
| "tie_word_embeddings": true |
| }, |
| "audio_end_token_id": 250036, |
| "audio_start_token_id": 250035, |
| "audio_token_id": 250062, |
| "auto_map": { |
| "AutoConfig": "configuration_jais_asr.JaisASRConfig", |
| "AutoModel": "modeling_jais_asr.JaisASRModel", |
| "AutoModelForCausalLM": "modeling_jais_asr.JaisASRForConditionalGeneration", |
| "AutoProcessor": "processing_jais_asr.JaisASRProcessor" |
| }, |
| "bos_token_id": 0, |
| "custom_pipelines": { |
| "automatic-speech-recognition": { |
| "impl": "pipeline_jais_asr.JaisASRTranscriptionPipeline", |
| "pt": [ |
| "AutoModelForCausalLM" |
| ], |
| "tf": [] |
| } |
| }, |
| "dtype": "bfloat16", |
| "eos_token_id": 1, |
| "model_type": "jais_asr", |
| "pad_token_id": 0, |
| "projector_hidden_act": "gelu", |
| "quantization_config": { |
| "config_groups": { |
| "group_0": { |
| "format": "float-quantized", |
| "input_activations": { |
| "actorder": null, |
| "block_structure": null, |
| "dynamic": true, |
| "group_size": 128, |
| "num_bits": 8, |
| "observer": null, |
| "observer_kwargs": {}, |
| "scale_dtype": null, |
| "strategy": "group", |
| "symmetric": true, |
| "type": "float", |
| "zp_dtype": null |
| }, |
| "output_activations": null, |
| "targets": [ |
| "Linear" |
| ], |
| "weights": { |
| "actorder": null, |
| "block_structure": [ |
| 128, |
| 128 |
| ], |
| "dynamic": false, |
| "group_size": null, |
| "num_bits": 8, |
| "observer": "memoryless_minmax", |
| "observer_kwargs": {}, |
| "scale_dtype": null, |
| "strategy": "block", |
| "symmetric": true, |
| "type": "float", |
| "zp_dtype": null |
| } |
| } |
| }, |
| "format": "float-quantized", |
| "global_compression_ratio": null, |
| "ignore": [ |
| "model.audio_tower.layers.0.self_attn.k_proj", |
| "model.audio_tower.layers.0.self_attn.v_proj", |
| "model.audio_tower.layers.0.self_attn.q_proj", |
| "model.audio_tower.layers.0.self_attn.out_proj", |
| "model.audio_tower.layers.0.fc1", |
| "model.audio_tower.layers.0.fc2", |
| "model.audio_tower.layers.1.self_attn.k_proj", |
| "model.audio_tower.layers.1.self_attn.v_proj", |
| "model.audio_tower.layers.1.self_attn.q_proj", |
| "model.audio_tower.layers.1.self_attn.out_proj", |
| "model.audio_tower.layers.1.fc1", |
| "model.audio_tower.layers.1.fc2", |
| "model.audio_tower.layers.2.self_attn.k_proj", |
| "model.audio_tower.layers.2.self_attn.v_proj", |
| "model.audio_tower.layers.2.self_attn.q_proj", |
| "model.audio_tower.layers.2.self_attn.out_proj", |
| "model.audio_tower.layers.2.fc1", |
| "model.audio_tower.layers.2.fc2", |
| "model.audio_tower.layers.3.self_attn.k_proj", |
| "model.audio_tower.layers.3.self_attn.v_proj", |
| "model.audio_tower.layers.3.self_attn.q_proj", |
| "model.audio_tower.layers.3.self_attn.out_proj", |
| "model.audio_tower.layers.3.fc1", |
| "model.audio_tower.layers.3.fc2", |
| "model.audio_tower.layers.4.self_attn.k_proj", |
| "model.audio_tower.layers.4.self_attn.v_proj", |
| "model.audio_tower.layers.4.self_attn.q_proj", |
| "model.audio_tower.layers.4.self_attn.out_proj", |
| "model.audio_tower.layers.4.fc1", |
| "model.audio_tower.layers.4.fc2", |
| "model.audio_tower.layers.5.self_attn.k_proj", |
| "model.audio_tower.layers.5.self_attn.v_proj", |
| "model.audio_tower.layers.5.self_attn.q_proj", |
| "model.audio_tower.layers.5.self_attn.out_proj", |
| "model.audio_tower.layers.5.fc1", |
| "model.audio_tower.layers.5.fc2", |
| "model.audio_tower.layers.6.self_attn.k_proj", |
| "model.audio_tower.layers.6.self_attn.v_proj", |
| "model.audio_tower.layers.6.self_attn.q_proj", |
| "model.audio_tower.layers.6.self_attn.out_proj", |
| "model.audio_tower.layers.6.fc1", |
| "model.audio_tower.layers.6.fc2", |
| "model.audio_tower.layers.7.self_attn.k_proj", |
| "model.audio_tower.layers.7.self_attn.v_proj", |
| "model.audio_tower.layers.7.self_attn.q_proj", |
| "model.audio_tower.layers.7.self_attn.out_proj", |
| "model.audio_tower.layers.7.fc1", |
| "model.audio_tower.layers.7.fc2", |
| "model.audio_tower.layers.8.self_attn.k_proj", |
| "model.audio_tower.layers.8.self_attn.v_proj", |
| "model.audio_tower.layers.8.self_attn.q_proj", |
| "model.audio_tower.layers.8.self_attn.out_proj", |
| "model.audio_tower.layers.8.fc1", |
| "model.audio_tower.layers.8.fc2", |
| "model.audio_tower.layers.9.self_attn.k_proj", |
| "model.audio_tower.layers.9.self_attn.v_proj", |
| "model.audio_tower.layers.9.self_attn.q_proj", |
| "model.audio_tower.layers.9.self_attn.out_proj", |
| "model.audio_tower.layers.9.fc1", |
| "model.audio_tower.layers.9.fc2", |
| "model.audio_tower.layers.10.self_attn.k_proj", |
| "model.audio_tower.layers.10.self_attn.v_proj", |
| "model.audio_tower.layers.10.self_attn.q_proj", |
| "model.audio_tower.layers.10.self_attn.out_proj", |
| "model.audio_tower.layers.10.fc1", |
| "model.audio_tower.layers.10.fc2", |
| "model.audio_tower.layers.11.self_attn.k_proj", |
| "model.audio_tower.layers.11.self_attn.v_proj", |
| "model.audio_tower.layers.11.self_attn.q_proj", |
| "model.audio_tower.layers.11.self_attn.out_proj", |
| "model.audio_tower.layers.11.fc1", |
| "model.audio_tower.layers.11.fc2", |
| "model.audio_tower.layers.12.self_attn.k_proj", |
| "model.audio_tower.layers.12.self_attn.v_proj", |
| "model.audio_tower.layers.12.self_attn.q_proj", |
| "model.audio_tower.layers.12.self_attn.out_proj", |
| "model.audio_tower.layers.12.fc1", |
| "model.audio_tower.layers.12.fc2", |
| "model.audio_tower.layers.13.self_attn.k_proj", |
| "model.audio_tower.layers.13.self_attn.v_proj", |
| "model.audio_tower.layers.13.self_attn.q_proj", |
| "model.audio_tower.layers.13.self_attn.out_proj", |
| "model.audio_tower.layers.13.fc1", |
| "model.audio_tower.layers.13.fc2", |
| "model.audio_tower.layers.14.self_attn.k_proj", |
| "model.audio_tower.layers.14.self_attn.v_proj", |
| "model.audio_tower.layers.14.self_attn.q_proj", |
| "model.audio_tower.layers.14.self_attn.out_proj", |
| "model.audio_tower.layers.14.fc1", |
| "model.audio_tower.layers.14.fc2", |
| "model.audio_tower.layers.15.self_attn.k_proj", |
| "model.audio_tower.layers.15.self_attn.v_proj", |
| "model.audio_tower.layers.15.self_attn.q_proj", |
| "model.audio_tower.layers.15.self_attn.out_proj", |
| "model.audio_tower.layers.15.fc1", |
| "model.audio_tower.layers.15.fc2", |
| "model.audio_tower.layers.16.self_attn.k_proj", |
| "model.audio_tower.layers.16.self_attn.v_proj", |
| "model.audio_tower.layers.16.self_attn.q_proj", |
| "model.audio_tower.layers.16.self_attn.out_proj", |
| "model.audio_tower.layers.16.fc1", |
| "model.audio_tower.layers.16.fc2", |
| "model.audio_tower.layers.17.self_attn.k_proj", |
| "model.audio_tower.layers.17.self_attn.v_proj", |
| "model.audio_tower.layers.17.self_attn.q_proj", |
| "model.audio_tower.layers.17.self_attn.out_proj", |
| "model.audio_tower.layers.17.fc1", |
| "model.audio_tower.layers.17.fc2", |
| "model.audio_tower.layers.18.self_attn.k_proj", |
| "model.audio_tower.layers.18.self_attn.v_proj", |
| "model.audio_tower.layers.18.self_attn.q_proj", |
| "model.audio_tower.layers.18.self_attn.out_proj", |
| "model.audio_tower.layers.18.fc1", |
| "model.audio_tower.layers.18.fc2", |
| "model.audio_tower.layers.19.self_attn.k_proj", |
| "model.audio_tower.layers.19.self_attn.v_proj", |
| "model.audio_tower.layers.19.self_attn.q_proj", |
| "model.audio_tower.layers.19.self_attn.out_proj", |
| "model.audio_tower.layers.19.fc1", |
| "model.audio_tower.layers.19.fc2", |
| "model.audio_tower.layers.20.self_attn.k_proj", |
| "model.audio_tower.layers.20.self_attn.v_proj", |
| "model.audio_tower.layers.20.self_attn.q_proj", |
| "model.audio_tower.layers.20.self_attn.out_proj", |
| "model.audio_tower.layers.20.fc1", |
| "model.audio_tower.layers.20.fc2", |
| "model.audio_tower.layers.21.self_attn.k_proj", |
| "model.audio_tower.layers.21.self_attn.v_proj", |
| "model.audio_tower.layers.21.self_attn.q_proj", |
| "model.audio_tower.layers.21.self_attn.out_proj", |
| "model.audio_tower.layers.21.fc1", |
| "model.audio_tower.layers.21.fc2", |
| "model.audio_tower.layers.22.self_attn.k_proj", |
| "model.audio_tower.layers.22.self_attn.v_proj", |
| "model.audio_tower.layers.22.self_attn.q_proj", |
| "model.audio_tower.layers.22.self_attn.out_proj", |
| "model.audio_tower.layers.22.fc1", |
| "model.audio_tower.layers.22.fc2", |
| "model.audio_tower.layers.23.self_attn.k_proj", |
| "model.audio_tower.layers.23.self_attn.v_proj", |
| "model.audio_tower.layers.23.self_attn.q_proj", |
| "model.audio_tower.layers.23.self_attn.out_proj", |
| "model.audio_tower.layers.23.fc1", |
| "model.audio_tower.layers.23.fc2", |
| "model.audio_tower.layers.24.self_attn.k_proj", |
| "model.audio_tower.layers.24.self_attn.v_proj", |
| "model.audio_tower.layers.24.self_attn.q_proj", |
| "model.audio_tower.layers.24.self_attn.out_proj", |
| "model.audio_tower.layers.24.fc1", |
| "model.audio_tower.layers.24.fc2", |
| "model.audio_tower.layers.25.self_attn.k_proj", |
| "model.audio_tower.layers.25.self_attn.v_proj", |
| "model.audio_tower.layers.25.self_attn.q_proj", |
| "model.audio_tower.layers.25.self_attn.out_proj", |
| "model.audio_tower.layers.25.fc1", |
| "model.audio_tower.layers.25.fc2", |
| "model.audio_tower.layers.26.self_attn.k_proj", |
| "model.audio_tower.layers.26.self_attn.v_proj", |
| "model.audio_tower.layers.26.self_attn.q_proj", |
| "model.audio_tower.layers.26.self_attn.out_proj", |
| "model.audio_tower.layers.26.fc1", |
| "model.audio_tower.layers.26.fc2", |
| "model.audio_tower.layers.27.self_attn.k_proj", |
| "model.audio_tower.layers.27.self_attn.v_proj", |
| "model.audio_tower.layers.27.self_attn.q_proj", |
| "model.audio_tower.layers.27.self_attn.out_proj", |
| "model.audio_tower.layers.27.fc1", |
| "model.audio_tower.layers.27.fc2", |
| "model.audio_tower.layers.28.self_attn.k_proj", |
| "model.audio_tower.layers.28.self_attn.v_proj", |
| "model.audio_tower.layers.28.self_attn.q_proj", |
| "model.audio_tower.layers.28.self_attn.out_proj", |
| "model.audio_tower.layers.28.fc1", |
| "model.audio_tower.layers.28.fc2", |
| "model.audio_tower.layers.29.self_attn.k_proj", |
| "model.audio_tower.layers.29.self_attn.v_proj", |
| "model.audio_tower.layers.29.self_attn.q_proj", |
| "model.audio_tower.layers.29.self_attn.out_proj", |
| "model.audio_tower.layers.29.fc1", |
| "model.audio_tower.layers.29.fc2", |
| "model.audio_tower.layers.30.self_attn.k_proj", |
| "model.audio_tower.layers.30.self_attn.v_proj", |
| "model.audio_tower.layers.30.self_attn.q_proj", |
| "model.audio_tower.layers.30.self_attn.out_proj", |
| "model.audio_tower.layers.30.fc1", |
| "model.audio_tower.layers.30.fc2", |
| "model.audio_tower.layers.31.self_attn.k_proj", |
| "model.audio_tower.layers.31.self_attn.v_proj", |
| "model.audio_tower.layers.31.self_attn.q_proj", |
| "model.audio_tower.layers.31.self_attn.out_proj", |
| "model.audio_tower.layers.31.fc1", |
| "model.audio_tower.layers.31.fc2", |
| "model.audio_tower.conv_out", |
| "model.audio_tower.proj1", |
| "model.audio_tower.proj2", |
| "model.embed_audio.linear_1", |
| "model.embed_audio.linear_2", |
| "lm_head" |
| ], |
| "kv_cache_scheme": null, |
| "quant_method": "compressed-tensors", |
| "quantization_status": "compressed", |
| "sparsity_config": {}, |
| "transform_config": {}, |
| "version": "0.18.0" |
| }, |
| "text_config": { |
| "attention_bias": false, |
| "attention_dropout": 0.0, |
| "attention_gate_func": null, |
| "bos_token_id": 0, |
| "dtype": "bfloat16", |
| "eos_token_id": 1, |
| "head_dim": 128, |
| "hidden_act": "silu", |
| "hidden_size": 2560, |
| "initializer_range": 0.02, |
| "intermediate_size": 10240, |
| "layernorm_num_groups": 2, |
| "max_position_embeddings": 524288, |
| "mlp_only_layers": [ |
| 0, |
| 1, |
| 2, |
| 3, |
| 4, |
| 5, |
| 6, |
| 7, |
| 8, |
| 9, |
| 10, |
| 11, |
| 12, |
| 13, |
| 14, |
| 15, |
| 16, |
| 17, |
| 18, |
| 19, |
| 20, |
| 21, |
| 22, |
| 23, |
| 24, |
| 25, |
| 26, |
| 27, |
| 28, |
| 29, |
| 30, |
| 31, |
| 32, |
| 33, |
| 34, |
| 35 |
| ], |
| "model_type": "k2_horizon", |
| "moe_intermediate_size": 0, |
| "mova_num_experts": 0, |
| "mova_num_experts_per_tok": 0, |
| "num_attention_heads": 32, |
| "num_experts": 0, |
| "num_experts_per_tok": 0, |
| "num_hidden_layers": 36, |
| "num_key_value_heads": 8, |
| "num_shared_experts": 0, |
| "pad_token_id": 0, |
| "query_key_norm": false, |
| "rms_norm_eps": 1e-06, |
| "rope_head_dim": 128, |
| "rope_parameters": { |
| "rope_theta": 10000000.0, |
| "rope_type": "default" |
| }, |
| "rope_theta": 10000000.0, |
| "sliding_window": null, |
| "tie_word_embeddings": false, |
| "use_cache": true, |
| "use_sliding_window": false, |
| "vocab_size": 250624 |
| }, |
| "tie_word_embeddings": false, |
| "transformers_version": "5.13.0" |
| } |
