Update training graphs to 2x2 layout with speed tracking tile#72
Merged
Conversation
Change plot_training_curves from 1x3 to 2x2 grid across all three species notebooks. The new layout shows Eval Reward, Episode Length, a species-specific diagnostic (Tilt Angle for velociraptor, Pelvis Height for T-Rex and brachiosaurus), and Forward Velocity (Speed). For T-Rex and brachiosaurus, add diagnostics.npz persistence to DiagnosticsCallback (log_dir parameter, _save_diagnostics method, per-rollout history tracking) so forward_vel and other INFO_KEYS are available for plotting. Also add completed_stages accumulator and incremental plot_training_curves calls after each stage. https://claude.ai/code/session_01XsJBwgKtKLsS36FBw9ZP4o
Bring both notebooks to full parity with velociraptor's logging and diagnostics infrastructure: - DiagnosticsCallback: add termination reason tracking (Counter), correct species-specific REWARD_KEYS and INFO_KEYS to match what each environment actually emits - train_stage: replace evaluate_policy with evaluate_with_forward_vel, add curriculum gating (reward + episode length + forward velocity thresholds), add write_stage_summary per-stage text output, enrich stage_results dict with episode length, forward vel, sim_dt, gate info - make_env / create_vec_env: add log_dir parameter for Monitor file output - record_stage_video: add vecnorm_path for normalized obs during recording - plot_training_curves: add curriculum threshold dashed lines, switch bottom-left panel from Pelvis Height to Tilt Angle (now tracked) - write_training_summary / save_results_json: include episode length, forward velocity, and sim time in output - Stage cells: pass vecnorm_path to record_stage_video - Section 10 video loop: pass vecnorm_path per stage https://claude.ai/code/session_01XsJBwgKtKLsS36FBw9ZP4o
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This pull request updates the
notebooks/brachiosaurus_training.ipynbnotebook to improve reproducibility and streamline the process for recording videos and plotting training curves during curriculum training. The changes ensure that vector normalization files are consistently passed when recording stage videos and that training curves are automatically updated and saved after each stage.Enhancements to video recording and training curve plotting:
record_stage_videocalls to include thevecnorm_pathparameter, ensuring correct normalization is used for each stage. [1] [2] [3] [4]completed_stagesand callsplot_training_curvesto save updated training curves, improving workflow automation. [1] [2] [3]plot_training_curvefunction with a single call toplot_training_curves, simplifying the code and ensuring consistency.These changes make the notebook easier to use and maintain, reducing manual steps and potential errors when visualizing training progress.