| { | |
| "_name_or_path": "checkpoints/namo-500m-native-pixelshuffle_2x/checkpoint-15476", | |
| "architectures": [ | |
| "NamoForCausalLM" | |
| ], | |
| "audio_config": { | |
| "model_type": "whisper" | |
| }, | |
| "conn_ve_llm_lr": 1e-05, | |
| "conn_ve_llm_type": "pixelshuffle_2x", | |
| "dynamic_size": true, | |
| "freeze_conn_ve_llm": false, | |
| "ignore_index": -100, | |
| "image_aspect_ratio": "pad", | |
| "image_seq_length": 576, | |
| "image_token_index": -200, | |
| "longest_edge": 700, | |
| "mm_use_im_patch_token": false, | |
| "mm_use_im_start_end": false, | |
| "model_type": "namo", | |
| "multimodal_projector_bias": true, | |
| "new_img_size": null, | |
| "s2": false, | |
| "s2_max_split_size": 384, | |
| "s2_scales": "384,768", | |
| "shortest_edge": null, | |
| "text_config": { | |
| "_attn_implementation_autoset": true, | |
| "_name_or_path": "checkpoints/Qwen2.5-0.5B-Instruct", | |
| "architectures": [ | |
| "Qwen2ForCausalLM" | |
| ], | |
| "bos_token_id": 151643, | |
| "eos_token_id": 151645, | |
| "hidden_size": 896, | |
| "intermediate_size": 4864, | |
| "max_window_layers": 21, | |
| "model_type": "qwen2", | |
| "num_attention_heads": 14, | |
| "num_hidden_layers": 24, | |
| "num_key_value_heads": 2, | |
| "rope_theta": 1000000.0, | |
| "tie_word_embeddings": true, | |
| "torch_dtype": "bfloat16" | |
| }, | |
| "tokenizer_model_max_length": 131072, | |
| "tokenizer_padding_side": "right", | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.48.1", | |
| "tune_conn_ve_llm": false, | |
| "unfreeze_ve": true, | |
| "use_cache": false, | |
| "video_fps": 1, | |
| "video_frames_num": 16, | |
| "vision_config": { | |
| "_name_or_path": "checkpoints/aimv2-l-native-trained-base", | |
| "architectures": [ | |
| "AIMv2Model" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_aimv2.AIMv2Config", | |
| "AutoModel": "modeling_aimv2.AIMv2Model", | |
| "FlaxAutoModel": "modeling_flax_aimv2.FlaxAIMv2Model" | |
| }, | |
| "image_size": 224, | |
| "intermediate_size": 2816, | |
| "model_type": "aimv2", | |
| "num_attention_heads": 8, | |
| "num_queries": 256, | |
| "projection_dropout": 0.0, | |
| "qkv_bias": false, | |
| "rms_norm_eps": 1e-05, | |
| "torch_dtype": "bfloat16", | |
| "use_bias": false | |
| }, | |
| "vision_feature_layer": -2, | |
| "vision_feature_select_strategy": "same" | |
| } | |