Upload 4 files

cubecomposer-3k/args.json

@@ -0,0 +1 @@
+{"max_pixels": 921600, "height": 1600, "width": 3200, "num_frames": 27, "data_file_keys": "image,video", "dataset_repeat": 1, "model_paths": null, "model_id_with_origin_paths": "Wan-AI/Wan2.2-TI2V-5B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-TI2V-5B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-TI2V-5B:Wan2.2_VAE.pth", "learning_rate": 1e-05, "num_epochs": 100, "output_path": "/group/40075/lingenli/exp-outputs/20251103_024747-CubeComposer-Wan2.2-5B-Full-FoV60to120-Fr27Win3-MultiRes512to800-HistoryWin2-FragAdjacent-DiagContext-32Padding-FRBLUD-CrossPosEmbsPad-FaceWisePrompts-GlobalSinkToken", "remove_prefix_in_ckpt": "pipe.dit.", "trainable_models": "dit", "lora_base_model": null, "lora_target_modules": "q,k,v,o,ffn.0,ffn.2", "lora_rank": 32, "lora_checkpoint": null, "extra_inputs": null, "use_gradient_checkpointing_offload": false, "gradient_accumulation_steps": 1, "max_timestep_boundary": 1.0, "min_timestep_boundary": 0.0, "find_unused_parameters": false, "save_steps": null, "dataset_num_workers": 0, "weight_decay": 0.01, "csv_path": "/group/40005/public_datasets/WEB360/WEB360/WEB360_360TF_train.csv", "video_dir": "/group/40005/public_datasets/WEB360/WEB360/videos_512x1024x100", "yt_caption_dir": "/group/40075/public_datasets/youtube_360_caption/captions", "yt_video_dir": "/group/40075/public_datasets/youtube_360_caption/extracted_clips", "yt_format": "mp4", "odv_root_dir": "/group/40005/public_datasets/ODVista360", "padding_width": 32, "window_length": 9, "window_stride": 9, "cube_map_size": 768, "resume_from_ckpt": "/group/40075/lingenli/exp-outputs/20251030_022907-CubeComposer-Wan2.2-5B-Full-FoV60to120-Fr27Win3-MultiRes512to768-HistoryWin2-FragAdjacent-DiagContext-32Padding-FRBLUD-CrossPosEmbsPad-FaceWisePrompts/epoch-41.safetensors", "validation_interval_steps": 200, "active_faces": "F,R,B,L,U,D", "fuse_vae_embedding_in_latents": false, "use_gradient_checkpointing": true, "use_vanilla_pos_embs": false, "condition_mode": "diagonal-context", "max_history_windows": 2, "fragment_future_context": true, "face_prompt_prob": 0.5, "use_face_prompts_in_val": true, "use_random_fov": true, "use_random_num_waypoints": true, "inference_boundary_padding": true, "inference_boundary_pixel_blending": true, "enable_multi_resolution_training": true, "min_resolution": 512, "max_resolution": 800, "base_model": "wan2.2", "inference_simulation_training": false, "inference_simulation_min_steps": 5, "inference_simulation_max_steps": 15, "inference_simulation_grad_accum": 15, "inference_simulation_loss_weighting": false, "inference_simulation_loss_weight_scale": 2.0, "seperated_timestep_modulation": true, "use_global_sink_token": true}

cubecomposer-3k/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:473576ce7d0feb38bfd002ab555629337163b3f2c9178cb81f566e1d4c6822a2
+size 10002031752

cubecomposer-4k/args.json

@@ -0,0 +1 @@
+{"max_pixels": 921600, "height": 1920, "width": 3840, "num_frames": 15, "data_file_keys": "image,video", "dataset_repeat": 1, "model_paths": null, "model_id_with_origin_paths": "Wan-AI/Wan2.2-TI2V-5B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-TI2V-5B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-TI2V-5B:Wan2.2_VAE.pth", "learning_rate": 1e-05, "num_epochs": 100, "output_path": "/group/40075/lingenli/exp-outputs/20251104_195815-CubeComposer-Wan2.2-5B-Full-FoV60to120-Fr15Win3-Res4K-HistoryWin2-FragAdjacent-DiagContext-32Padding-FRBLUD-CrossPosEmbsPad-FaceWisePrompts-GlobalSinkToken-LatentStorage", "remove_prefix_in_ckpt": "pipe.dit.", "trainable_models": "dit", "lora_base_model": null, "lora_target_modules": "q,k,v,o,ffn.0,ffn.2", "lora_rank": 32, "lora_checkpoint": null, "extra_inputs": null, "use_gradient_checkpointing_offload": false, "gradient_accumulation_steps": 1, "max_timestep_boundary": 1.0, "min_timestep_boundary": 0.0, "find_unused_parameters": false, "save_steps": null, "dataset_num_workers": 0, "weight_decay": 0.01, "csv_path": "/group/40005/public_datasets/WEB360/WEB360/WEB360_360TF_train.csv", "video_dir": "/group/40005/public_datasets/WEB360/WEB360/videos_512x1024x100", "yt_caption_dir": "/group/40075/public_datasets/youtube_360_caption/captions", "yt_video_dir": "/group/40075/public_datasets/youtube_360_caption/extracted_clips", "yt_format": "mp4", "odv_root_dir": "/group/40005/public_datasets/ODVista360", "padding_width": 32, "window_length": 5, "window_stride": 5, "cube_map_size": 960, "resume_from_ckpt": "/group/40005/lingenli/exp-outputs/20251103_024747-CubeComposer-Wan2.2-5B-Full-FoV60to120-Fr27Win3-MultiRes512to800-HistoryWin2-FragAdjacent-DiagContext-32Padding-FRBLUD-CrossPosEmbsPad-FaceWisePrompts-GlobalSinkToken/epoch-6.safetensors", "validation_interval_steps": 500, "validation_num_inference_steps": 15, "active_faces": "F,R,B,L,U,D", "fuse_vae_embedding_in_latents": false, "use_gradient_checkpointing": true, "use_vanilla_pos_embs": false, "condition_mode": "diagonal-context", "max_history_windows": 2, "fragment_future_context": true, "face_prompt_prob": 0.5, "use_face_prompts_in_val": true, "use_random_fov": true, "use_random_num_waypoints": true, "inference_boundary_padding": true, "inference_boundary_pixel_blending": true, "enable_multi_resolution_training": false, "min_resolution": 960, "max_resolution": 960, "base_model": "wan2.2", "inference_simulation_training": false, "inference_simulation_min_steps": 5, "inference_simulation_max_steps": 15, "inference_simulation_grad_accum": 15, "inference_simulation_loss_weighting": false, "inference_simulation_loss_weight_scale": 2.0, "seperated_timestep_modulation": true, "use_global_sink_token": true}

cubecomposer-4k/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:efe4d7c5f60dd879624b1b5a9c1b41a99ee2d686c33f3102e139ffd82ebabd39
+size 10002031752