{ "model_type": "phiva", "quantization": "q4f16_0", "model_config": { "image_token_index": 32011, "text_config": { "model_type": "llava", "hidden_size": 3072, "vocab_size": 32013, "num_hidden_layers": 32, "num_attention_heads": 32, "intermediate_size": 8192, "rms_norm_eps": 1e-06, "num_key_value_heads": 32, "position_embedding_base": 10000, "context_window_size": 131072, "prefill_chunk_size": 2048, "head_dim": 96, "tensor_parallel_shards": 1, "max_batch_size": 1, "kwargs": { "_name_or_path": "phiva-4b-pretrained", "architectures": [ "LlavaForConditionalGeneration" ], "embd_pdrop": 0.0, "eos_token_id": 32000, "original_max_position_embeddings": 4096, "pad_token_id": 32000, "resid_pdrop": 0.0, "sliding_window": null, "torch_dtype": "float16", "ignore_index": -100, "image_token_index": 32011, "projector_hidden_act": "gelu", "text_config": { "_name_or_path": "microsoft/Phi-3-mini-4k-instruct", "architectures": [ "Phi3ForCausalLM" ], "embd_pdrop": 0.0, "eos_token_id": 32000, "hidden_size": 3072, "intermediate_size": 8192, "num_attention_heads": 32, "num_hidden_layers": 32, "num_key_value_heads": 32, "max_position_embeddings": 131072, "model_type": "phi3", "original_max_position_embeddings": 4096, "pad_token_id": 32000, "resid_pdrop": 0.0, "rms_norm_eps": 1e-06, "sliding_window": 2047, "torch_dtype": "float16", "vocab_size": 32013 }, "tie_word_embeddings": false, "transformers_version": "4.42.3", "vision_config": { "hidden_size": 768, "image_size": 224, "intermediate_size": 3072, "model_type": "clip_vision_model", "num_attention_heads": 12, "num_hidden_layers": 12, "patch_size": 16, "projection_dim": 512, "vocab_size": 32013 }, "vision_feature_layer": -2, "vision_feature_select_strategy": "default" } }, "vision_config": { "hidden_size": 768, "image_size": 224, "intermediate_size": 3072, "num_attention_heads": 12, "num_hidden_layers": 12, "patch_size": 16, "projection_dim": 512, "vocab_size": 32013, "num_channels": 3, "layer_norm_eps": 1e-06, "kwargs": { "model_type": "clip_vision_model" } }, "vocab_size": 32013, "context_window_size": 768, "sliding_window_size": -1, "prefill_chunk_size": 2048, "tensor_parallel_shards": 1, "max_batch_size": 80, "text_architecture": "LlavaForConditionalGeneration" }, "vocab_size": 32013, "context_window_size": 768, "sliding_window_size": -1, "prefill_chunk_size": 2048, "attention_sink_size": -1, "tensor_parallel_shards": 1, "mean_gen_len": 128, "max_gen_len": 512, "shift_fill_factor": 0.3, "temperature": 0, "presence_penalty": 0.0, "frequency_penalty": 0.0, "repetition_penalty": 1.0, "top_p": 0.95, "conv_template": { "name": "phiva", "system_template": "{system_message}", "system_message": "", "add_role_after_system_message": false, "roles": { "user": "USER", "assistant": "ASSISTANT" }, "role_templates": { "user": "{user_message}", "assistant": "{assistant_message}", "tool": "{tool_message}" }, "messages": [], "seps": [ " " ], "role_content_sep": ": ", "role_empty_sep": ":", "stop_str": [ "" ], "stop_token_ids": [ 32000 ], "function_string": "", "use_function_calling": false, "image_token_index": 32011 }, "pad_token_id": 32000, "bos_token_id": 1, "eos_token_id": 32000, "tokenizer_files": [ "tokenizer.model", "tokenizer.json", "added_tokens.json", "tokenizer_config.json" ], "version": "0.1.0" }