WishArdently/InternVideo2Stage2-VisionEncoder
1128
1{2 "architectures": [3 "InternVideo2Stage2VideoEncoder"4 ],5 "auto_map": {6 "AutoConfig": "config.InternVideo2Config",7 "AutoModel": "model.InternVideo2Stage2VideoEncoder"8 },9 "auto_resume": false,10 "batch_size": 64,11 "batch_size_test": 4,12 "best_key": [13 "msrvtt_1k_test_match",14 "t2v_r1"15 ],16 "compile_model": false,17 "criterion": {18 "clip_loss_ratio": [19 1.0,20 1.021 ],22 "distill_final_features": true,23 "loss_weight": {24 "mlm": 1.0,25 "mvm": 0.0,26 "uta": 0.0,27 "vtc": 1.0,28 "vtm": 1.029 },30 "mlm_masking_prob": 0.5,31 "vtm_hard_neg": true32 },33 "debug": false,34 "deep_fusion": false,35 "deepspeed": {36 "enable": true,37 "stage": 138 },39 "delete_ds_optim_states": true,40 "device": "cuda",41 "dist_url": "env://",42 "evaluate": false,43 "evaluation": {44 "eval_frame_ensemble": "concat",45 "eval_offload": true,46 "eval_x_only": false,47 "k_test": 12848 },49 "gradient_checkpointing": true,50 "inputs": {51 "batch_size": {52 "image": 64,53 "video": 6454 },55 "batch_size_test": {56 "image": 4,57 "video": 458 },59 "image_res": 224,60 "max_txt_l": {61 "image": 32,62 "video": 3263 },64 "video_input": {65 "num_frames": 8,66 "num_frames_test": 8,67 "random_aug": false,68 "sample_type": "rand",69 "sample_type_test": "middle"70 }71 },72 "jump_evaluate": false,73 "log_freq": 100,74 "max_txt_l": 32,75 "mode": "pt",76 "model": {77 "embed_dim": 512,78 "find_unused_parameters": false,79 "model_cls": "InternVideo2_Stage2",80 "multimodal": {81 "enable": true82 },83 "temp": 0.07,84 "text_encoder": "bert_large",85 "vision_encoder": {86 "checkpoint_num": 40,87 "clip_embed_dim": 768,88 "clip_input_resolution": 224,89 "clip_norm_type": "l2",90 "clip_return_layer": 6,91 "clip_student_return_interval": 1,92 "clip_teacher": null,93 "clip_teacher_embed_dim": 3200,94 "clip_teacher_final_dim": 768,95 "clip_teacher_return_interval": 1,96 "d_model": 1408,97 "image_mask_ratio": 0.5,98 "image_mask_type": "random",99 "img_size": 224,100 "keep_temporal": false,101 "name": "pretrain_internvideo2_1b_patch14_224",102 "num_frames": 8,103 "only_mask": true,104 "patch_size": 14,105 "pretrained": "/home/linanxi/InternVideo/checkpoints/InternVideo2-stage2_1b-224p-f4/InternVideo2-stage2_1b-224p-f4.pt",106 "sep_image_video_pos_embed": true,107 "tubelet_size": 1,108 "use_checkpoint": false,109 "use_flash_attn": true,110 "use_fused_mlp": true,111 "use_fused_rmsnorm": true,112 "video_mask_ratio": 0.8,113 "video_mask_type": "random"114 }115 },116 "model_type": "internvideo2",117 "num_frames": 8,118 "num_frames_test": 8,119 "num_workers": 6,120 "optimizer": {121 "different_lr": {122 "enable": false,123 "lr": 0.001,124 "module_names": []125 },126 "lr": 5e-05,127 "max_grad_norm": 3.0,128 "opt": "adamW",129 "opt_betas": [130 0.9,131 0.98132 ],133 "weight_decay": 0.05134 },135 "output_dir": null,136 "pretrained_path": "",137 "resume": false,138 "save_ckpt_iter": null,139 "save_latest": true,140 "scheduler": {141 "epochs": 10,142 "min_lr_multi": 0.01,143 "sched": "cosine",144 "warmup_epochs": 1145 },146 "seed": 42,147 "test_file": {148 "didemo_ret_test": "available_corpus[\"didemo_ret_test\"]",149 "msrvtt_1k_test": "available_corpus[\"msrvtt_1k_test\"]"150 },151 "test_types": [152 "msrvtt_1k_test",153 "didemo_ret_test"154 ],155 "text_enc": "bert_large",156 "tokenizer": null,157 "torch_dtype": "float16",158 "train_file": "available_corpus[\"pretrain_example_data_1B\"]",159 "transformers_version": "4.47.0",160 "use_bf16": true,161 "use_flash_sdp": false,162 "use_half_precision": false,163 "use_mem_efficient_sdp": false,164 "wandb": {165 "enable": false,166 "entity": "opengvlab",167 "project": "InternVideo2-Stage2"168 }169}170 