{ "architectures": [ "GraniteSpeechForConditionalGeneration" ], "audio_token_index": 100352, "downsample_rate": 1, "dtype": "float32", "encoder_config": { "context_size": 30, "conv_expansion_factor": 2, "conv_kernel_size": 3, "dim_head": 16, "dropout": 0.1, "dtype": "bfloat16", "feedforward_mult": 2, "hidden_dim": 32, "input_dim": 16, "max_pos_emb": 64, "model_type": "granite_speech_encoder", "num_heads": 2, "num_layers": 2, "output_dim": 32 }, "has_lora_adapter": false, "initializer_range": 0.02, "model_type": "granite_speech", "projector_config": { "_attn_implementation_autoset": true, "attention_probs_dropout_prob": 0.1, "cross_attention_frequency": 1, "dtype": "bfloat16", "encoder_hidden_size": 32, "hidden_act": "gelu", "hidden_dropout_prob": 0.1, "hidden_size": 32, "initializer_range": 0.02, "intermediate_size": 64, "layer_norm_eps": 1e-12, "max_position_embeddings": 2048, "model_type": "blip_2_qformer", "num_attention_heads": 2, "num_hidden_layers": 1, "pad_token_id": 0, "position_embedding_type": "absolute", "use_qformer_text_input": false, "vocab_size": 30522 }, "text_config": { "_name_or_path": "/proj/speech/saon/slam-llm/29.2-c/granite-4.0-1b-base", "architectures": [ "GraniteForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "attention_multiplier": 0.0078125, "bos_token_id": 100257, "dtype": "float32", "embedding_multiplier": 12, "eos_token_id": 100257, "head_dim": 16, "hidden_act": "silu", "hidden_size": 64, "initializer_range": 0.1, "intermediate_size": 32, "logits_scaling": 8, "max_position_embeddings": 4096, "mlp_bias": false, "model_type": "granite", "num_attention_heads": 4, "num_hidden_layers": 2, "num_key_value_heads": 2, "pad_token_id": 100256, "residual_multiplier": 0.22, "rms_norm_eps": 1e-05, "rope_parameters": { "rope_theta": 10000, "rope_type": "default" }, "tie_word_embeddings": false, "use_cache": true, "vocab_size": 100353 }, "tie_word_embeddings": false, "transformers.js_config": { "use_external_data_format": { "audio_encoder.onnx": 1, "decoder_model_merged.onnx": 1, "embed_tokens.onnx": 1, "audio_encoder_fp16.onnx": 1, "decoder_model_merged_fp16.onnx": 1, "embed_tokens_fp16.onnx": 1 }, "kv_cache_dtype": { "q4f16": "float16", "fp16": "float16" } }, "transformers_version": "5.3.0.dev0", "window_size": 3 }