Production Testing Updates & Fixes - #1
Draft
Exios66 wants to merge 12 commits into
Draft
Conversation
…d README for setup instructions, added W&B configuration in training scripts, and included logging of metrics during training. Updated requirements to include wandb package.
… Changed tokenizer to PreTrainedTokenizerFast, adjusted n_positions to accommodate longer sequences, and enhanced data preparation by loading from parquet files. Added functionality to flatten multiline states and updated evaluation loss reporting.
…data processing and model training scripts, streamlined README for clarity, and enhanced .gitignore to exclude generated artifacts. Updated requirements to include additional packages for improved functionality.
Document a full set of comparable pokerAI experiments (model, architecture, optimization, tokenizer, data, imbalance, systems, inference) with configs, commands, and success criteria. Add scripts/run_experiment.py plus sweep YAMLs and honor WANDB_NAME in trainers for named runs. Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Jack J Burleson // LJM <Exios66@users.noreply.github.com>
* Fix critical pipeline bugs across data, training, and inference Address separator mismatch, missing cleaned data from convert, mandatory W&B crashes, TRL CPU bf16 failure, BOS/EOS and truncation misalignment, offset-based action masking, tokenizer corpus mismatch, and stale docs. Co-authored-by: Jack J Burleson // LJM <Exios66@users.noreply.github.com> * Align experiment catalog with optional W&B and comma data format Document that W&B logging is optional, fix EXP-E3 separator guidance now that convert uses to_line commas, and make run_experiment/sweeps announce or require W&B so catalog runs do not silently skip tracking. Co-authored-by: Jack J Burleson // LJM <Exios66@users.noreply.github.com> --------- Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Jack J Burleson // LJM <Exios66@users.noreply.github.com>
* Improve data handling by addressing separator issues and ensuring cleaned data is included in conversion processes. * Update experiment catalog documentation to clarify optional W&B logging and correct separator guidance for new data formats. * Fix various bugs related to training and inference, including action masking and tokenizer mismatches. Co-authored-by: Jack J Burleson // LJM <Exios66@users.noreply.github.com> Co-authored-by: Cursor Agent <cursoragent@cursor.com>
* Change W&B project and entity names to `mooslin-university-of-wisconsin-madison/poker-ai` for consistency across documentation and scripts. * Modify training scripts to ensure proper W&B project initialization and logging. * Update README and experiment catalog to reflect new W&B project details and improve clarity on logging configurations. Co-authored-by: Jack J Burleson // LJM <Exios66@users.noreply.github.com> Co-authored-by: Cursor Agent <cursoragent@cursor.com>
…es (#4) * Fix critical pipeline bugs across data, training, and inference Address separator mismatch, missing cleaned data from convert, mandatory W&B crashes, TRL CPU bf16 failure, BOS/EOS and truncation misalignment, offset-based action masking, tokenizer corpus mismatch, and stale docs. Co-authored-by: Jack J Burleson // LJM <Exios66@users.noreply.github.com> * Align experiment catalog with optional W&B and comma data format Document that W&B logging is optional, fix EXP-E3 separator guidance now that convert uses to_line commas, and make run_experiment/sweeps announce or require W&B so catalog runs do not silently skip tracking. Co-authored-by: Jack J Burleson // LJM <Exios66@users.noreply.github.com> * Add action-aware eval charts and alternative experiment approaches Extend the CLI with majority, structured-feature (RF/LogReg), and class-weighted GPT-2 trainers. Add per-action metrics, confusion/F1 charts, occlusion feature importance, and scripts/evaluate.py for post-hoc W&B visualization. Co-authored-by: Jack J Burleson // LJM <Exios66@users.noreply.github.com> --------- Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Jack J Burleson // LJM <Exios66@users.noreply.github.com>
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.
Patches & Enhancements Encountered in Testing
July 11th, 2026