Repository navigation
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
jadenfix
changed the title
This was referenced Oct 13, 2026
This branch had an error being deployed
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Repair the policy constructor keyword and YAML learning-rate conversion, mask terminal bootstrapping, and preserve singleton loss shapes. Real TTS now fails explicitly when unavailable; synthetic audio requires TTSConfig(allow_mock=True), and synthesis failures raise instead of returning silence. Honor the configured speech model name.
Replace the stale system test that mixed incompatible DQN/PPO APIs with a real local replay/optimizer/checkpoint workflow. It collects 12 simulated transitions, performs 9 DQN updates, checks changed parameters and checkpoint equality, then runs inference.
Validation: all 27 tests pass with model downloads disabled. This includes explicit synthetic WAV plumbing and fail-closed speech errors. Real speech still requires Coqui TTS and pretrained weights; the environment itself retains placeholder observations/rewards and is not a live conversation workflow. Launcher/CI repairs remain separate in #1.