Uniform quantile data processing#24
Open
T4ras123 wants to merge 3 commits into
Open
Conversation
added 3 commits
February 27, 2026 01:41
- Introduced to compute the binning range R from training data, including radius proxy calculations and overflow counts. - Added to fit bin configurations (uniform and quantile) from training data, producing JSON files for later use in encoding and decoding. - Implemented class for unified configuration management of binning methods, enhancing the encoding/decoding process. - Updated data preprocessing to support new binning methods and configurations, ensuring compatibility with existing workflows.
- Removed extensive docstrings from , , and to streamline the code and enhance readability. - Kept function signatures intact while eliminating unnecessary comments, focusing on maintaining functionality and clarity.
- Introduced new paths for revisited cartesian and quantile binned datasets in paths.yaml. - Added Qwen3 tokenizer configuration files, including metadata, tokenizer settings, and a chat template for the binned_258 variant. - Updated count_tokens.py to remove an unnecessary parameter in the tokenizer loading process, enhancing clarity.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This pull request introduces new scripts for computing coordinate ranges and fitting bin configurations, and integrates support for bin configuration objects in the data preprocessing pipeline. The main goal is to provide more robust and flexible handling of coordinate binning, including quantile-based and uniform binning, and to facilitate downstream usage of these bin configurations for encoding molecular coordinates.
New scripts for data analysis and bin fitting:
scripts/compute_range_R.pyfor computing recommended coordinate ranges based on conformer radius proxies and quantile statistics from the training set. This script helps determine appropriate value ranges for coordinate binning and reports overflow statistics for train/val/test splits.scripts/fit_bins.pyfor fitting uniform and quantile bin configurations to molecular coordinate data. The script pools coordinates, computes distribution summaries, fits bins, saves bin configs as JSON, and reports overflow statistics for each split.Bin configuration integration and preprocessing improvements:
quantile_bins.json,uniform_bins.json) for use with coordinate encoding routines. These files specify bin edges, ranges, and metadata for downstream encoding. [1] [2]read_moland_read_mol_implindata_preprocessing.pyto accept an optionalbin_configargument and support the newencode_cartesian_with_configembedding function, enabling flexible binning based on external configuration. [1] [2] [3]Configuration updates:
paths.yamlto add an additional checkpoint path forqwen_yerevann_rootand to update the model root forqw600_pre_binned_paired. [1] [2][Copilot is generating a summary...]