|
{ |
|
"architectures": [ |
|
"TinyLlavaForConditionalGeneration" |
|
], |
|
"cache_dir": null, |
|
"connector_type": "linear", |
|
"hidden_size": 3584, |
|
"ignore_index": -100, |
|
"image_aspect_ratio": "square", |
|
"image_token_index": -200, |
|
"llm_model_name_or_path": "Qwen/Qwen2.5-7B", |
|
"model_type": "tinyllava", |
|
"num_queries": 128, |
|
"num_resampler_layers": 3, |
|
"pad_token": "<|endoftext|>", |
|
"pad_token_id": 151643, |
|
"resampler_hidden_size": 768, |
|
"text_config": { |
|
"_name_or_path": "Qwen/Qwen2.5-7B", |
|
"architectures": [ |
|
"Qwen2ForCausalLM" |
|
], |
|
"bos_token_id": 151643, |
|
"eos_token_id": 151643, |
|
"hidden_size": 3584, |
|
"intermediate_size": 18944, |
|
"max_position_embeddings": 131072, |
|
"max_window_layers": 28, |
|
"model_type": "qwen2", |
|
"num_attention_heads": 28, |
|
"num_hidden_layers": 28, |
|
"num_key_value_heads": 4, |
|
"rope_theta": 1000000.0, |
|
"sliding_window": null, |
|
"torch_dtype": "bfloat16", |
|
"use_mrope": false, |
|
"use_sliding_window": false, |
|
"vocab_size": 152064 |
|
}, |
|
"tokenizer_model_max_length": 2048, |
|
"tokenizer_name_or_path": "Qwen/Qwen2.5-7B", |
|
"tokenizer_padding_side": "right", |
|
"tokenizer_use_fast": false, |
|
"torch_dtype": "bfloat16", |
|
"transformers_version": "4.45.1", |
|
"tune_type_connector": "full", |
|
"tune_type_llm": "full", |
|
"tune_type_vision_tower": "frozen", |
|
"tune_vision_tower_from_layer": 0, |
|
"use_cache": true, |
|
"vision_config": { |
|
"hidden_act": "gelu", |
|
"hidden_size": 768, |
|
"image_size": 512, |
|
"intermediate_size": 3072, |
|
"model_name_or_path": "StanfordAIMI/XrayCLIP__vit-b-16__laion2b-s34b-b88k", |
|
"model_name_or_path2": "", |
|
"model_type": "clip_vision_model", |
|
"num_attention_heads": 12, |
|
"num_hidden_layers": 12, |
|
"patch_size": 16, |
|
"projection_dim": 512 |
|
}, |
|
"vision_feature_layer": -2, |
|
"vision_feature_select_strategy": "patch", |
|
"vision_hidden_size": 768, |
|
"vision_model_name_or_path": "StanfordAIMI/XrayCLIP__vit-b-16__laion2b-s34b-b88k", |
|
"vision_model_name_or_path2": "", |
|
"vocab_size": 152064 |
|
} |
|
|