Upload folder using huggingface_hub

config.json

@@ -13,8 +13,8 @@
   "do_mask_loss_for_padding": false,
   "down_dims": [
     512,
-    1024,
-    2048
+    512,
+    512
   ],
   "horizon": 16,
   "input_normalization_modes": {
@@ -46,7 +46,7 @@
       6
     ]
   },
-  "prediction_type": "epsilon",
+  "prediction_type": "sample",
   "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
   "spatial_softmax_num_keypoints": 32,
   "use_film_scale_modulation": true,

config.yaml

@@ -2,7 +2,7 @@ resume: false
 device: cuda
 use_amp: false
 seed: 1000
-dataset_repo_id: jmercat/koch_feed_cat
+dataset_repo_id: jmercat/koch_feed_cat_2
 video_backend: pyav
 training:
   offline_steps: 16000
@@ -123,8 +123,8 @@ policy:
   spatial_softmax_num_keypoints: 32
   down_dims:
   - 512
-  - 1024
-  - 2048
+  - 512
+  - 512
   kernel_size: 5
   n_groups: 8
   diffusion_step_embed_dim: 128
@@ -134,7 +134,7 @@ policy:
   beta_schedule: squaredcos_cap_v2
   beta_start: 0.0001
   beta_end: 0.02
-  prediction_type: epsilon
+  prediction_type: sample
   clip_sample: true
   clip_sample_range: 1.0
   num_inference_steps: null

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:74715073bae6f24078e59f423a53528c311bdccea9702b2834a9f7d632e14664
-size 1043856648
+oid sha256:f24b5cb5a19c9e198a386282287488c3f06bf6d21e8e298765f095f8f4c0ee08
+size 210612832