tangger/lerobot
1
0
Fork 1
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Use PytorchModelHubMixin to save models as safetensors (#125)

Browse Source 
Co-authored-by: Remi <re.cadene@gmail.com>
This commit is contained in:
Alexander Soare
2024-05-01 16:17:18 +01:00
committed by GitHub
parent 01d5490d44
commit a4891095e4
18 changed files with 556 additions and 527 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
lerobot/common/policies/act/configuration_act.py
View File

@@ -38,7 +38,7 @@ class ACTConfig:
replace_final_stride_with_dilation: Whether to replace the ResNet's final 2x2 stride with a dilated
convolution.
pre_norm: Whether to use "pre-norm" in the transformer blocks.
d_model: The transformer blocks' main hidden dimension.
dim_model: The transformer blocks' main hidden dimension.
n_heads: The number of heads to use in the transformer blocks' multi-head attention.
dim_feedforward: The dimension to expand the transformer's hidden dimension to in the feed-forward
layers.
@@ -94,7 +94,7 @@ class ACTConfig:
replace_final_stride_with_dilation: int = False
# Transformer layers.
pre_norm: bool = False
d_model: int = 512
dim_model: int = 512
n_heads: int = 8
dim_feedforward: int = 3200
feedforward_activation: str = "relu"