{ "model": { "type": "gemma4", "vocab_size": 262144, "context_length": 131072, "decoder": { "session_options": { "log_id": "onnxruntime-genai", "provider_options": [ { "RyzenAI": { "hybrid_opt_max_seq_length": "4096", "hybrid_opt_npu_jit_order": "runtime", "external_data_file": "decoder/model.pb.bin", "hybrid_opt_token_backend": "npu" } } ] }, "filename": "decoder/model.onnx", "head_size": 256, "hidden_size": 2560, "inputs": { "inputs_embeds": "inputs_embeds", "attention_mask": "attention_mask", "per_layer_inputs": "per_layer_inputs", "past_key_names": "past_key_values.%d.key", "past_value_names": "past_key_values.%d.value" }, "outputs": { "logits": "logits", "present_key_names": "present.%d.key", "present_value_names": "present.%d.value" }, "num_attention_heads": 8, "num_hidden_layers": 42, "num_key_value_heads": 2 }, "bos_token_id": 2, "eos_token_id": [ 1, 106 ], "pad_token_id": 0, "vision": { "filename": "vision_encoder/model.onnx", "config_filename": "image_processor.json", "inputs": { "pixel_values": "pixel_values", "pixel_position_ids": "pixel_position_ids" }, "outputs": { "image_features": "image_features" }, "session_options": { "log_id": "onnxruntime-genai", "provider_options": [] }, "spatial_merge_size": 2 }, "embedding": { "filename": "embedding/model.onnx", "inputs": { "input_ids": "input_ids", "image_features": "image_features", "audio_features": "audio_features" }, "outputs": { "inputs_embeds": "inputs_embeds", "per_layer_inputs": "per_layer_inputs" }, "session_options": { "log_id": "onnxruntime-genai", "provider_options": [] } }, "speech": { "filename": "audio_encoder/model.onnx", "config_filename": "audio_feature_extraction.json", "inputs": { "audio_embeds": "input_features", "attention_mask": "input_features_mask" }, "outputs": { "audio_features": "audio_features_Q4" }, "session_options": { "log_id": "onnxruntime-genai", "provider_options": [] } }, "image_token_id": 258880, "audio_token_id": 258881, "boa_token_id": 256000 }, "search": { "do_sample": true, "early_stopping": true, "max_length": 131072, "min_length": 0, "num_beams": 1, "num_return_sequences": 1, "past_present_share_buffer": true, "repetition_penalty": 1.0, "temperature": 1.0, "top_k": 1, "top_p": 1.0 } }