Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Make diffusion work#6

Merged
Cadene merged 7 commits intomainfrom
user/rcadene/2024_03_04_diffusion
Mar 4, 2024
Merged

Make diffusion work#6
Cadene merged 7 commits intomainfrom
user/rcadene/2024_03_04_diffusion

Conversation

@Cadene
Copy link
Contributor

@Cadene Cadene commented Mar 4, 2024

  • add normalization
    • add automatic mean_std normalization
    • add automatic min_max normalization
    • add manually set normalization (will do in future PR)
  • add queue of observations during rollout to access n previous images and agent positions
  • add access to n previous images and agent positions during update of diffusion policy
  • use pusht + diffusion in default.yaml
  • log num parameters and offline dataset size

@Cadene Cadene requested a review from aliberts March 4, 2024 13:28
@Cadene Cadene changed the title [WIP] Make diffusion work Make diffusion work Mar 4, 2024
@Cadene Cadene merged commit e990f3e into main Mar 4, 2024
@aliberts aliberts deleted the user/rcadene/2024_03_04_diffusion branch March 9, 2024 15:07
alexander-soare pushed a commit that referenced this pull request Jun 5, 2024
Merge from hf lerobot main branch
menhguin pushed a commit to menhguin/lerobot that referenced this pull request Feb 9, 2025
brysonjones pushed a commit to brysonjones/lerobot that referenced this pull request Nov 12, 2025
southfreebird pushed a commit to acting-machines/lerobot that referenced this pull request Feb 1, 2026
* add config for eval and fix eval from config; provided model should give avg_max_reward=0.6341779769630806

* add script to visualise reward distribution from eval

* add docstring

* add max reward to wandb

* add script to fit and visualsise actions tokeniser and add option to policy config

* restore drop_n_last_frames to 1

* default fast_tokenizer_path

* PR comment fixes + add plot coeffs
southfreebird pushed a commit to acting-machines/lerobot that referenced this pull request Feb 1, 2026
* add config for eval and fix eval from config; provided model should give avg_max_reward=0.6341779769630806

* add script to visualise reward distribution from eval

* add docstring

* add max reward to wandb

* add script to fit and visualsise actions tokeniser and add option to policy config

* restore drop_n_last_frames to 1

* default fast_tokenizer_path

* PR comment fixes + add plot coeffs
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant