SmolVLA training configs and SPA episode viewer#339
Closed
vertix wants to merge 5 commits intoPositronic-Robotics:mainfrom
Closed
SmolVLA training configs and SPA episode viewer#339vertix wants to merge 5 commits intoPositronic-Robotics:mainfrom
vertix wants to merge 5 commits intoPositronic-Robotics:mainfrom
Conversation
- Align `scheduler_decay_steps` with total training steps by default, preventing LR from hitting floor early in long runs - Add `jinja2` to dependencies (required by starlette's Jinja2Templates) - Remove `full_report=True` from sim eval `StackingSuccess` observer
- Fix SmolVLAConfig import (was XVLAConfig) - Add explicit `batch_size=64` (SmolVLA recommended default) - Extract `_train()` to support multiple training configs - Add `full_finetune` config (unfreezes vision encoder + all params) - Multi-config CLI: `train` and `full_finetune` subcommands
Use rerun viewer's `add_receiver`/`remove_receiver` API to swap RRD
data without reloading the iframe. Adds Service Worker to cache WASM
assets and `/api/episode/{id}` endpoint for client-side navigation.
vertix
changed the title
- Reset `viewerReady` on iframe fallback reload - Use `replaceState` in popstate handler to avoid double history entries - Restore `full_report=True` in sim config (unintentional removal)
Contributor
Author
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
SmolVLAConfiginstead ofXVLAConfig), set recommendedbatch_size=64, fix LR scheduler decay, addjinja2depfull_finetunetraining config (unfreezes vision encoder + all params) alongside default expert-onlytrainadd_receiver/remove_receiverAPI to swap RRD data without reloading the 35MB WASM. Adds Service Worker for WASM caching and/api/episode/{id}JSON endpointTest plan
trainandfull_finetuneon sim_stack dataset (60K steps each on H100)