Upload policy weights, train config and readme

README.md

@@ -1,22 +1,21 @@
 ---
-base_model: lerobot/smolvla_base
-datasets: data/merged_multitask
+datasets: data/banana_plate
 library_name: lerobot
 license: apache-2.0
-model_name: smolvla
+model_name: diffusion
 pipeline_tag: robotics
 tags:
 - lerobot
 - robotics
-- smolvla
+- diffusion
 ---
 
-# Model Card for smolvla
+# Model Card for diffusion
 
 <!-- Provide a quick summary of what the model is/does. -->
 
 
-[SmolVLA](https://huggingface.co/papers/2506.01844) is a compact, efficient vision-language-action model that achieves competitive performance at reduced computational costs and can be deployed on consumer-grade hardware.
+[Diffusion Policy](https://huggingface.co/papers/2303.04137) treats visuomotor control as a generative diffusion process, producing smooth, multi-step action trajectories that excel at contact-rich manipulation.
 
 
 This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).

config.json

@@ -1,6 +1,6 @@
 {
-    "type": "smolvla",
-    "n_obs_steps": 1,
+    "type": "diffusion",
+    "n_obs_steps": 2,
     "input_features": {
         "observation.state": {
             "type": "STATE",
@@ -43,55 +43,57 @@
     },
     "device": "cuda",
     "use_amp": false,
+    "use_peft": false,
     "push_to_hub": true,
     "repo_id": "False",
     "private": null,
     "tags": null,
     "license": null,
     "pretrained_path": null,
-    "chunk_size": 50,
-    "n_action_steps": 50,
+    "horizon": 16,
+    "n_action_steps": 8,
     "normalization_mapping": {
-        "VISUAL": "IDENTITY",
-        "STATE": "MEAN_STD",
-        "ACTION": "MEAN_STD"
+        "VISUAL": "MEAN_STD",
+        "STATE": "MIN_MAX",
+        "ACTION": "MIN_MAX"
     },
-    "max_state_dim": 32,
-    "max_action_dim": 32,
-    "resize_imgs_with_padding": [
+    "drop_n_last_frames": 7,
+    "vision_backbone": "resnet18",
+    "crop_shape": [
+        84,
+        84
+    ],
+    "crop_is_random": true,
+    "pretrained_backbone_weights": null,
+    "use_group_norm": true,
+    "spatial_softmax_num_keypoints": 32,
+    "use_separate_rgb_encoder_per_camera": false,
+    "down_dims": [
         512,
-        512
+        1024,
+        2048
     ],
-    "empty_cameras": 0,
-    "adapt_to_pi_aloha": false,
-    "use_delta_joint_actions_aloha": false,
-    "tokenizer_max_length": 48,
-    "num_steps": 10,
-    "use_cache": true,
-    "freeze_vision_encoder": true,
-    "train_expert_only": true,
-    "train_state_proj": true,
+    "kernel_size": 5,
+    "n_groups": 8,
+    "diffusion_step_embed_dim": 128,
+    "use_film_scale_modulation": true,
+    "noise_scheduler_type": "DDPM",
+    "num_train_timesteps": 100,
+    "beta_schedule": "squaredcos_cap_v2",
+    "beta_start": 0.0001,
+    "beta_end": 0.02,
+    "prediction_type": "epsilon",
+    "clip_sample": true,
+    "clip_sample_range": 1.0,
+    "num_inference_steps": null,
+    "do_mask_loss_for_padding": false,
     "optimizer_lr": 0.0001,
     "optimizer_betas": [
-        0.9,
-        0.95
+        0.95,
+        0.999
     ],
     "optimizer_eps": 1e-08,
-    "optimizer_weight_decay": 1e-10,
-    "optimizer_grad_clip_norm": 10,
-    "scheduler_warmup_steps": 1000,
-    "scheduler_decay_steps": 30000,
-    "scheduler_decay_lr": 2.5e-06,
-    "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
-    "load_vlm_weights": false,
-    "add_image_special_tokens": false,
-    "attention_mode": "cross_attn",
-    "prefix_length": -1,
-    "pad_language_to": "longest",
-    "num_expert_layers": -1,
-    "num_vlm_layers": 16,
-    "self_attn_every_n_layers": 2,
-    "expert_width_multiplier": 0.75,
-    "min_period": 0.004,
-    "max_period": 4.0
+    "optimizer_weight_decay": 1e-06,
+    "scheduler_name": "cosine",
+    "scheduler_warmup_steps": 500
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8378d98d2f003917a142cf2bc75628f6a3adb04be96ef1822f149bd5436282e9
-size 1197789224
+oid sha256:5ab9433255460c5696acd8c2c0a083c57bc4903f60d8f31f3c93342a97f5ebd0
+size 1082658752

train_config.json

@@ -1,6 +1,6 @@
 {
     "dataset": {
-        "repo_id": "data/merged_multitask",
+        "repo_id": "data/banana_plate",
         "root": null,
         "episodes": null,
         "image_transforms": {
@@ -81,8 +81,8 @@
     },
     "env": null,
     "policy": {
-        "type": "smolvla",
-        "n_obs_steps": 1,
+        "type": "diffusion",
+        "n_obs_steps": 2,
         "input_features": {
             "observation.state": {
                 "type": "STATE",
@@ -125,60 +125,62 @@
         },
         "device": "cuda",
         "use_amp": false,
+        "use_peft": false,
         "push_to_hub": true,
         "repo_id": "False",
         "private": null,
         "tags": null,
         "license": null,
         "pretrained_path": null,
-        "chunk_size": 50,
-        "n_action_steps": 50,
+        "horizon": 16,
+        "n_action_steps": 8,
         "normalization_mapping": {
-            "VISUAL": "IDENTITY",
-            "STATE": "MEAN_STD",
-            "ACTION": "MEAN_STD"
+            "VISUAL": "MEAN_STD",
+            "STATE": "MIN_MAX",
+            "ACTION": "MIN_MAX"
         },
-        "max_state_dim": 32,
-        "max_action_dim": 32,
-        "resize_imgs_with_padding": [
+        "drop_n_last_frames": 7,
+        "vision_backbone": "resnet18",
+        "crop_shape": [
+            84,
+            84
+        ],
+        "crop_is_random": true,
+        "pretrained_backbone_weights": null,
+        "use_group_norm": true,
+        "spatial_softmax_num_keypoints": 32,
+        "use_separate_rgb_encoder_per_camera": false,
+        "down_dims": [
             512,
-            512
+            1024,
+            2048
         ],
-        "empty_cameras": 0,
-        "adapt_to_pi_aloha": false,
-        "use_delta_joint_actions_aloha": false,
-        "tokenizer_max_length": 48,
-        "num_steps": 10,
-        "use_cache": true,
-        "freeze_vision_encoder": true,
-        "train_expert_only": true,
-        "train_state_proj": true,
+        "kernel_size": 5,
+        "n_groups": 8,
+        "diffusion_step_embed_dim": 128,
+        "use_film_scale_modulation": true,
+        "noise_scheduler_type": "DDPM",
+        "num_train_timesteps": 100,
+        "beta_schedule": "squaredcos_cap_v2",
+        "beta_start": 0.0001,
+        "beta_end": 0.02,
+        "prediction_type": "epsilon",
+        "clip_sample": true,
+        "clip_sample_range": 1.0,
+        "num_inference_steps": null,
+        "do_mask_loss_for_padding": false,
         "optimizer_lr": 0.0001,
         "optimizer_betas": [
-            0.9,
-            0.95
+            0.95,
+            0.999
         ],
         "optimizer_eps": 1e-08,
-        "optimizer_weight_decay": 1e-10,
-        "optimizer_grad_clip_norm": 10,
-        "scheduler_warmup_steps": 1000,
-        "scheduler_decay_steps": 30000,
-        "scheduler_decay_lr": 2.5e-06,
-        "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
-        "load_vlm_weights": false,
-        "add_image_special_tokens": false,
-        "attention_mode": "cross_attn",
-        "prefix_length": -1,
-        "pad_language_to": "longest",
-        "num_expert_layers": -1,
-        "num_vlm_layers": 16,
-        "self_attn_every_n_layers": 2,
-        "expert_width_multiplier": 0.75,
-        "min_period": 0.004,
-        "max_period": 4.0
+        "optimizer_weight_decay": 1e-06,
+        "scheduler_name": "cosine",
+        "scheduler_warmup_steps": 500
     },
-    "output_dir": "/media/choi/HDD/outputs/train/smolvla_merged",
-    "job_name": "smolvla_merged",
+    "output_dir": "/media/choi/HDD/outputs/train/Diffusion_banana_plate",
+    "job_name": "diffusion_merged",
     "resume": false,
     "seed": 1000,
     "num_workers": 4,
@@ -186,26 +188,25 @@
     "steps": 100000,
     "eval_freq": 20000,
     "log_freq": 200,
+    "tolerance_s": 0.0001,
     "save_checkpoint": true,
     "save_freq": 20000,
     "use_policy_training_preset": true,
     "optimizer": {
-        "type": "adamw",
+        "type": "adam",
         "lr": 0.0001,
-        "weight_decay": 1e-10,
-        "grad_clip_norm": 10,
+        "weight_decay": 1e-06,
+        "grad_clip_norm": 10.0,
         "betas": [
-            0.9,
-            0.95
+            0.95,
+            0.999
         ],
         "eps": 1e-08
     },
     "scheduler": {
-        "type": "cosine_decay_with_warmup",
-        "num_warmup_steps": 1000,
-        "num_decay_steps": 30000,
-        "peak_lr": 0.0001,
-        "decay_lr": 2.5e-06
+        "type": "diffuser",
+        "num_warmup_steps": 500,
+        "name": "cosine"
     },
     "eval": {
         "n_episodes": 50,
@@ -221,6 +222,12 @@
         "run_id": null,
         "mode": null
     },
-    "checkpoint_path": null,
-    "rename_map": {}
+    "peft": null,
+    "use_rabc": false,
+    "rabc_progress_path": null,
+    "rabc_kappa": 0.01,
+    "rabc_epsilon": 1e-06,
+    "rabc_head_mode": "sparse",
+    "rename_map": {},
+    "checkpoint_path": null
 }