wangzixiang
/
lerobot
mirror of https://github.com/huggingface/lerobot.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

remove redundant explanations

This commit is contained in:
jayLEE0301 2024-06-05 11:33:19 -04:00
parent 78da71c31b
commit 06d3eb40e1
1 changed files with 1 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
lerobot/common/policies/vqbet/modeling_vqbet.py
View File

@ -343,16 +343,11 @@ class VQBeTHead(nn.Module):
and the output dimension of `self.map_to_cbet_preds_bin` is `self.config.vqvae_groups * self.config.vqvae_n_embed`, where
`self.config.vqvae_groups` is number of RVQ layers, and
`self.config.vqvae_n_embed` is codebook size of RVQ.
if the agent select the code sequentially, we use self.map_to_cbet_preds_primary_bin and self.map_to_cbet_preds_secondary_bin instead of self._map_to_cbet_preds_bin.
self.map_to_cbet_preds_offset: output the predicted offsets for all the codes in all the layers.
The input dimension of ` self.map_to_cbet_preds_offset` is same with the output of GPT,
and the output dimension of ` self.map_to_cbet_preds_offset` is `self.config.vqvae_groups * self.config.vqvae_n_embed * config.action_chunk_size * config.output_shapes["action"][0]`, where
`self.config.vqvae_groups` is number of RVQ layers,
`self.config.vqvae_n_embed` is codebook size of RVQ,
`config.action_chunk_size is action chunk size of each token, and
`config.output_shapes["action"][0]` is the dimension of action
and the output dimension of ` self.map_to_cbet_preds_offset` is `self.config.vqvae_groups * self.config.vqvae_n_embed * config.action_chunk_size * config.output_shapes["action"][0]`
"""
super().__init__()