Upload 4 files

ACT_DP_multitask/checkpoints/mixture_multiview_v3_50_500/act_dp/dataset_stats.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1757f1c5b672ca8c87e84afd096934ae5571ef1e04830499c2e3bc3be58d788
+size 1049

ACT_DP_multitask/checkpoints/mixture_multiview_v3_50_500/act_dp/policy_config.json

@@ -0,0 +1,56 @@
+{
+    "lr": 2e-05,
+    "task_name": "bottle_adjust",
+    "is_wandb": false,
+    "save_epoch": 100,
+    "num_queries": 50,
+    "predict_frame": 0,
+    "history_step": 0,
+    "resize_rate": 1,
+    "image_downsample_rate": 1,
+    "temporal_downsample_rate": 1,
+    "image_height": 480,
+    "image_width": 640,
+    "kl_weight": 10,
+    "hidden_dim": 512,
+    "dim_feedforward": 3200,
+    "lr_backbone": 1e-05,
+    "backbone": "resnet18",
+    "enc_layers": 4,
+    "dec_layers": 7,
+    "nheads": 8,
+    "camera_names": [
+        "cam_high",
+        "cam_left_wrist",
+        "cam_right_wrist"
+    ],
+    "norm_type": "minmax",
+    "disable_vae_latent": true,
+    "disable_resnet": false,
+    "condition_type": "cross_attention",
+    "num_inference_steps": 10,
+    "num_train_timesteps": 100,
+    "prediction_type": "sample",
+    "loss_type": "l1",
+    "schedule_type": "DDIM",
+    "beta_schedule": "squaredcos_cap_v2",
+    "diffusion_timestep_type": "cat",
+    "attention_type": "v0",
+    "causal_mask": false,
+    "predict_only_last": false,
+    "share_decoder": false,
+    "tokenizer_model_temporal_rate": 8,
+    "tokenizer_model_spatial_rate": 16,
+    "tokenizer_model_name": "Cosmos-Tokenizer-DV8x16x16",
+    "prediction_weight": 1,
+    "imitate_weight": 1,
+    "token_dim": 6,
+    "patch_size": 5,
+    "token_pe_type": "learned",
+    "next_frame": false,
+    "pretrain": false,
+    "mae": false,
+    "segmentation": false,
+    "use_film": true,
+    "multitask": true
+}

ACT_DP_multitask/checkpoints/mixture_multiview_v3_50_500/act_dp/train_args_config.json

@@ -0,0 +1,69 @@
+{
+    "is_wandb": false,
+    "ckpt_dir": "checkpoints/mixture_multiview_v3_50_500/act_dp",
+    "policy_class": "ACT_diffusion",
+    "task_name": "bottle_adjust",
+    "head_camera_type": "D435",
+    "num_episodes": 100,
+    "train_ratio": 0.99,
+    "batch_size": 42,
+    "seed": 0,
+    "num_epochs": 500,
+    "lr": 2e-05,
+    "lr_schedule_type": "constant",
+    "backbone": "resnet18",
+    "kl_weight": 10,
+    "save_epoch": 100,
+    "chunk_size": 50,
+    "history_step": 0,
+    "predict_frame": 0,
+    "image_downsample_rate": 1,
+    "resize_rate": 1,
+    "temporal_downsample_rate": 1,
+    "predict_only_last": false,
+    "disable_multi_view": false,
+    "hidden_dim": 512,
+    "dim_feedforward": 3200,
+    "norm_type": "minmax",
+    "num_train_steps": 100,
+    "num_inference_steps": 10,
+    "imitate_weight": 1,
+    "schedule_type": "DDIM",
+    "prediction_type": "sample",
+    "beta_schedule": "squaredcos_cap_v2",
+    "loss_type": "l1",
+    "diffusion_timestep_type": "cat",
+    "attention_type": "v0",
+    "causal_mask": false,
+    "disable_vae_latent": true,
+    "disable_resnet": false,
+    "disable_scale": false,
+    "share_decoder": false,
+    "condition_type": "cross_attention",
+    "tokenizer_model_type": "DV",
+    "tokenizer_model_temporal_rate": 8,
+    "tokenizer_model_spatial_rate": 16,
+    "prediction_weight": 1,
+    "patch_size": 5,
+    "token_pe_type": "learned",
+    "nf": false,
+    "pretrain": false,
+    "mae": false,
+    "seg": false,
+    "masks": false,
+    "resume": "",
+    "world_size": -1,
+    "rank": -1,
+    "dist_url": "tcp://224.66.41.62:23456",
+    "dist_backend": "nccl",
+    "gpu": 0,
+    "tokenizer_model_name": "Cosmos-Tokenizer-DV8x16x16",
+    "token_dim": 6,
+    "camera_names": [
+        "cam_high",
+        "cam_left_wrist",
+        "cam_right_wrist"
+    ],
+    "image_height": 240,
+    "image_width": 320
+}