Upload folder using huggingface_hub

README.md

@@ -1,10 +1,10 @@
 ---
 library_name: lerobot
 tags:
+- diffusion-policy
 - model_hub_mixin
 - pytorch_model_hub_mixin
 - robotics
-- vqbet
 ---
 
 This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration:

config.json

@@ -1,18 +1,22 @@
 {
-  "action_chunk_size": 5,
-  "bet_softmax_temperature": 0.1,
+  "beta_end": 0.02,
+  "beta_schedule": "squaredcos_cap_v2",
+  "beta_start": 0.0001,
+  "clip_sample": true,
+  "clip_sample_range": 1.0,
   "crop_is_random": true,
   "crop_shape": [
     84,
     84
   ],
-  "dropout": 0.1,
-  "gpt_block_size": 500,
-  "gpt_hidden_dim": 512,
-  "gpt_input_dim": 512,
-  "gpt_n_head": 8,
-  "gpt_n_layer": 8,
-  "gpt_output_dim": 512,
+  "diffusion_step_embed_dim": 128,
+  "do_mask_loss_for_padding": false,
+  "down_dims": [
+    512,
+    1024,
+    2048
+  ],
+  "horizon": 16,
   "input_normalization_modes": {
     "observation.image": "mean_std",
     "observation.state": "min_max"
@@ -27,11 +31,13 @@
       2
     ]
   },
-  "mlp_hidden_dim": 1024,
-  "n_action_pred_token": 7,
-  "n_obs_steps": 5,
-  "n_vqvae_training_steps": 20000,
-  "offset_loss_weight": 10000.0,
+  "kernel_size": 5,
+  "n_action_steps": 8,
+  "n_groups": 8,
+  "n_obs_steps": 2,
+  "noise_scheduler_type": "DDPM",
+  "num_inference_steps": null,
+  "num_train_timesteps": 100,
   "output_normalization_modes": {
     "action": "min_max"
   },
@@ -40,14 +46,11 @@
       2
     ]
   },
+  "prediction_type": "epsilon",
   "pretrained_backbone_weights": null,
-  "primary_code_loss_weight": 5.0,
-  "secondary_code_loss_weight": 0.5,
-  "sequentially_select": false,
   "spatial_softmax_num_keypoints": 32,
+  "use_film_scale_modulation": true,
   "use_group_norm": true,
-  "vision_backbone": "resnet18",
-  "vqvae_embedding_dim": 256,
-  "vqvae_enc_hidden_dim": 128,
-  "vqvae_n_embed": 16
+  "use_separate_rgb_encoder_per_camera": false,
+  "vision_backbone": "resnet18"
 }

config.yaml

@@ -59,30 +59,14 @@ training:
   - 0.999
   adam_eps: 1.0e-08
   adam_weight_decay: 1.0e-06
-  vqvae_lr: 0.001
-  n_vqvae_training_steps: 20000
-  bet_weight_decay: 0.0002
-  bet_learning_rate: 5.5e-05
-  bet_betas:
-  - 0.9
-  - 0.999
   delta_timestamps:
     observation.image:
-    - -0.4
-    - -0.3
-    - -0.2
     - -0.1
     - 0.0
     observation.state:
-    - -0.4
-    - -0.3
-    - -0.2
     - -0.1
     - 0.0
     action:
-    - -0.4
-    - -0.3
-    - -0.2
     - -0.1
     - 0.0
     - 0.1
@@ -95,6 +79,11 @@ training:
     - 0.8
     - 0.9
     - 1.0
+    - 1.1
+    - 1.2
+    - 1.3
+    - 1.4
+  drop_n_last_frames: 7
 eval:
   n_episodes: 50
   batch_size: 50
@@ -143,10 +132,10 @@ override_dataset_stats:
     - 511.0
     - 511.0
 policy:
-  name: vqbet
-  n_obs_steps: 5
-  n_action_pred_token: 7
-  action_chunk_size: 5
+  name: diffusion
+  n_obs_steps: 2
+  horizon: 16
+  n_action_steps: 8
   input_shapes:
     observation.image:
     - 3
@@ -170,20 +159,21 @@ policy:
   pretrained_backbone_weights: null
   use_group_norm: true
   spatial_softmax_num_keypoints: 32
-  n_vqvae_training_steps: ${training.n_vqvae_training_steps}
-  vqvae_n_embed: 16
-  vqvae_embedding_dim: 256
-  vqvae_enc_hidden_dim: 128
-  gpt_block_size: 500
-  gpt_input_dim: 512
-  gpt_output_dim: 512
-  gpt_n_layer: 8
-  gpt_n_head: 8
-  gpt_hidden_dim: 512
-  dropout: 0.1
-  mlp_hidden_dim: 1024
-  offset_loss_weight: 10000.0
-  primary_code_loss_weight: 5.0
-  secondary_code_loss_weight: 0.5
-  bet_softmax_temperature: 0.1
-  sequentially_select: false
+  down_dims:
+  - 512
+  - 1024
+  - 2048
+  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

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:670f97a37fb578a8da6eefa9ac8d019352899c0c6407e3a023433ce156c12a03
-size 158152442
+oid sha256:ebd4f2ba3500b315a4f45f6d7ff6c6acfd3891c768fbae945cdc17e7ceff5e76
+size 1050862408