Owner

mmcdermott

14 indexed repositories · View on GitHub

MEDS_transforms
A simple set of MEDS polars-based ETL and transformation functions

Python · 42 stars
MEDS_Tabular_AutoML
Limited automatic tabular ML pipelines for generic MEDS datasets.

Python · 20 stars
MEDS_EIC_AR
A MEDS, "Everything-is-code" style Autoregressive Generative Model, capable of zero-shot inference.

Python · 11 stars
nested_ragged_tensors
Utilities for efficiently working with, saving, and loading, collections of connected nested ragged tensors in PyTorch

Python · 9 stars
MEDS_extract
Helpers to aid in building ETLs for MEDS datasets leveraging the MEDS-Transforms library

Python · 9 stars
meds-torch-data
A package for creating pytorch dataset objects suitable for various modeling tasks from MEDS datasets.

Python · 5 stars
dftly
A simple library for a safe, expressive, config-file friendly, and readable DSL for encoding simple dataframe operations.

Python · 4 stars
ML_mixins
Useful Python Mixins for ML

Python · 3 stars
MEDS_model_template
A Copier template for building standards-conformant MEDS models with a mandated 5-step CLI (contributable to MEDS-DEV). ⚠️ AI-generated, ...

Python · 3 stars
MEDS_trajectory_evaluation
ACES Label Extraction Utilities for Zero-shot Generation based Prediction

Python · 3 stars
pretty-print-directory
A simple utility to pretty-print the contents of a directory

Python · 1 star
yaml_to_disk
A simple utility to define and populate a root directory on disk via a yaml or dictionary specification.

Python · 1 star
hiermark
Hierarchical, Annotatable Markdown (HAM): a 2D canvas of linked, editable markdown surfaces with rich annotations and block-anchored bran...

TypeScript · 1 star
MESSY_synth
Generate structurally-faithful synthetic raw data from any MEDS-Extract v0.7 MESSY config — for ETL demos and smoke tests.

Python · 0 stars

Beginner friendly Everything

Open All

label_trajectories crashes on a label window defined backward from a later endpoint Open

Difficulty 4/5 3-5 days Newbie friendliness 64/100

mmcdermott/MEDS_trajectory_evaluation#46 · Opened Sep 15, 2026
SGLang rolling generation crashes on zero-token RoPE tensor; possible empty prompts from finished rows Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

mmcdermott/MEDS_EIC_AR#176 · 1 comment · Opened Sep 15, 2026
Release workflow: changeset publish re-attempts already-published versions and exits 1 Open

Difficulty 3/5 1-2 days Newbie friendliness 58/100

mmcdermott/hiermark#103 · Opened Sep 7, 2026
Raise minimum SGLang version to avoid asyncio event-loop crash Open

Difficulty 2/5 1-3 hours Newbie friendliness 74/100

mmcdermott/MEDS_EIC_AR#175 · Opened Sep 7, 2026
label_trajectories corrupts window offsets when called repeatedly with the same config Open

Difficulty 2/5 1-3 hours Newbie friendliness 78/100

mmcdermott/MEDS_trajectory_evaluation#45 · Opened Sep 5, 2026
SGLang engine init raises the root log level to ERROR, silencing this repo's own warnings Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

mmcdermott/MEDS_EIC_AR#173 · Opened Sep 3, 2026
Grammar correctness test shouldn't depend on greedy decoding Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

mmcdermott/MEDS_EIC_AR#166 · Opened Sep 2, 2026
Make reshard_to_split's shard assignment a pure function of the subject (salted hash) rather than a cohort-wide permutation Open

Difficulty 5/5 Over a week Newbie friendliness 48/100

mmcdermott/MEDS_transforms#404 · Opened Sep 1, 2026
Allow multiple targets in _table.join (identity from one table, vocabulary from another) Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

mmcdermott/MEDS_extract#300 · Opened Aug 21, 2026
Make random subsequence sampling reproducible across seeded runs Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

mmcdermott/meds-torch-data#116 · Opened Aug 20, 2026
Value aggregations produce all-zero matrices due to suffixed code filter Open

Difficulty 2/5 1-3 hours Newbie friendliness 82/100

mmcdermott/MEDS_Tabular_AutoML#168 · Opened Aug 19, 2026
meds-tab-describe includes tuning and held-out shards when fitting feature metadata Open

Difficulty 4/5 3-5 days Newbie friendliness 52/100

mmcdermott/MEDS_Tabular_AutoML#166 · Opened Aug 19, 2026
SICdb run Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

mmcdermott/MESSY_synth#3 · Opened Aug 19, 2026
`source_block` changes default deduplication behavior in `merge_to_MEDS_cohort` Open

Difficulty 4/5 3-5 days Newbie friendliness 42/100

mmcdermott/MEDS_extract#299 · Opened Aug 8, 2026
Add a null test (is_null / is_present); today the only working idiom is `$x == $x` Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

mmcdermott/dftly#113 · Opened Aug 8, 2026
Cross-table projection widening can induce an unmodeled join-output collision Open

bug Parsing YAML Specification priority:low

Difficulty 4/5 3-5 days Newbie friendliness 45/100

mmcdermott/MEDS_extract#280 · Opened Aug 6, 2026
Single-argument function calls fail in string form: `?args` inlining hands `from_lark` a bare dict, read as kwargs Open

Difficulty 3/5 1-2 days Newbie friendliness 76/100

mmcdermott/dftly#109 · Opened Aug 5, 2026
Canonical 8-stage sequence is duplicated in EtlConfig.DEFAULT_PIPELINE, configs/_extract.yaml, and example/pipeline.yaml with no consistency check Open

Code Quality & Technical Debt priority:low Testing

Difficulty 2/5 1-3 hours Newbie friendliness 76/100

mmcdermott/MEDS_extract#239 · Opened Aug 5, 2026
Ctrl-C mid-child SIGKILLs the child 0.25s after SIGINT, truncating the download layer's engineered graceful teardown Open

bug priority:medium

Difficulty 4/5 3-5 days Newbie friendliness 55/100

mmcdermott/MEDS_extract#238 · Opened Aug 5, 2026
Registry entry naming a non-importable module escapes as a raw ModuleNotFoundError; docs falsely claim the registration is never imported Open

bug Logging / Technical Interface priority:low

Difficulty 3/5 1-2 days Newbie friendliness 74/100

mmcdermott/MEDS_extract#234 · Opened Aug 5, 2026
Targeted coverage gaps on risky error/concurrency paths (fail-fast abort, tar member types, PhysioNet 403 gate, fallback copies, RunConfig validation) Open

priority:medium Testing

Difficulty 4/5 3-5 days Newbie friendliness 48/100

mmcdermott/MEDS_extract#231 · Opened Aug 5, 2026
Stage-example framework comparison holes: lists compared as sets, non-comparable suffixes silently skipped, missing-output branch untested Open

bug priority:medium Testing

Difficulty 4/5 3-5 days Newbie friendliness 72/100

mmcdermott/MEDS_extract#229 · Opened Aug 5, 2026
Tiny-cohort split adjustment only repairs one empty split, raising even when a valid assignment exists Open

bug Extraction Stages priority:low

Difficulty 2/5 1-3 hours Newbie friendliness 78/100

mmcdermott/MEDS_extract#224 · Opened Aug 5, 2026
Overlapping external splits violate MEDS's one-shard-per-subject invariant and write duplicate rows to subject_splits.parquet Open

bug Extraction Stages priority:high

Difficulty 4/5 3-5 days Newbie friendliness 48/100

mmcdermott/MEDS_extract#223 · Opened Aug 5, 2026
split_and_shard_subjects has no worker gate: N parallel workers each full-scan every table and concurrently rewrite .shards.json Open

bug Computational Performance Extraction Stages priority:medium

Difficulty 2/5 1-3 hours Newbie friendliness 86/100

mmcdermott/MEDS_extract#220 · Opened Aug 5, 2026
Download CLI surfaces manifest-time failures (bad credentials, missing fsspec root, network down) as raw Hydra tracebacks Open

bug Logging / Technical Interface priority:medium

Difficulty 2/5 1-3 hours Newbie friendliness 78/100

mmcdermott/MEDS_extract#219 · Opened Aug 5, 2026
Tar symlink linknames are validated against the extraction root instead of the member's directory, rejecting legitimate archives Open

bug priority:low

Difficulty 2/5 1-3 hours Newbie friendliness 78/100

mmcdermott/MEDS_extract#218 · Opened Aug 5, 2026
HTTP intra-run Range resume without If-Range can silently corrupt sha-less files Open

bug priority:low

Difficulty 4/5 3-5 days Newbie friendliness 48/100

mmcdermott/MEDS_extract#216 · Opened Aug 5, 2026
Download: a manifest row literally named 'X.part' is silently destroyed by row 'X' staging Open

bug priority:low

Difficulty 3/5 1-2 days Newbie friendliness 72/100

mmcdermott/MEDS_extract#215 · Opened Aug 5, 2026
Download: unarchive cleanup (the "auto" default) deletes the only skip witness — every rerun re-downloads and re-extracts the archive Open

bug Computational Performance priority:high

Difficulty 4/5 3-5 days Newbie friendliness 52/100

mmcdermott/MEDS_extract#212 · Opened Aug 5, 2026
Download: interrupted or failed archive extraction is never repaired — reruns skip on the verified archive and exit 0 with members missing Open

bug priority:high

Difficulty 4/5 3-5 days Newbie friendliness 52/100

mmcdermott/MEDS_extract#211 · Opened Aug 5, 2026
Silently wrong value columns: bare-word values matching source columns, and near-miss MEDS column typos, pass without diagnostic Open

enhancement Parsing YAML Specification priority:medium Usability

Difficulty 4/5 3-5 days Newbie friendliness 68/100

mmcdermott/MEDS_extract#209 · Opened Aug 5, 2026
Non-mapping table/event blocks crash with bare dict() internals errors naming no table, event, or key Open

bug Logging / Technical Interface Parsing YAML Specification priority:medium

Difficulty 3/5 1-2 days Newbie friendliness 74/100

mmcdermott/MEDS_extract#208 · Opened Aug 5, 2026
Interpolations inside the etl: block are never resolved — '${...}' flows literally into provenance or triggers a bogus version-mismatch error Open

bug Parsing YAML Specification priority:medium

Difficulty 3/5 1-2 days Newbie friendliness 72/100

mmcdermott/MEDS_extract#206 · Opened Aug 5, 2026
Event value columns named source_block or code_components silently clobber (or are clobbered by) extractor internals Open

bug Parsing YAML Specification priority:medium

Difficulty 2/5 1-3 hours Newbie friendliness 84/100

mmcdermott/MEDS_extract#205 · Opened Aug 5, 2026
source_columns() drops a derived column's own input when a _table.cols entry shadows its source column, crashing stages later Open

bug Parsing YAML Specification priority:high

Difficulty 4/5 3-5 days Newbie friendliness 55/100

mmcdermott/MEDS_extract#202 · Opened Aug 5, 2026
Crash-stranded .strings.tmp.parquet intermediates are consumed by source globs on resume: permanently crashing resume or silent row duplication Open

bug priority:high

Difficulty 4/5 3-5 days Newbie friendliness 48/100

mmcdermott/MEDS_extract#201 · 1 comment · Opened Aug 5, 2026
Cast chains need nested parens and intermediate casts: `(((x)::int)::year)::datetime` for a three-step pipeline Open

Difficulty 5/5 Over a week Newbie friendliness 45/100

mmcdermott/dftly#98 · 1 comment · Opened Aug 4, 2026
MESSY: a table can declare only one `_table.join`, so lookup-then-aggregate-over-parent is inexpressible Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

mmcdermott/MEDS_extract#196 · Opened Aug 4, 2026
Aggregated-join advisory WARNING fires once per MessyConfig parse (8+ times per pipeline run) Open

Difficulty 3/5 1-2 days Newbie friendliness 55/100

mmcdermott/MEDS_extract#189 · Opened Aug 4, 2026
extract_code_metadata reducer computes the same frame three times (twice just for a log line) Open

Difficulty 2/5 1-3 hours Newbie friendliness 74/100

mmcdermott/MEDS_extract#188 · 2 comments · Opened Aug 4, 2026
EventConfig.extract eagerly collects a stats pass per event block, inside a lazy API Open

Difficulty 4/5 3-5 days Newbie friendliness 50/100

mmcdermott/MEDS_extract#187 · Opened Aug 4, 2026
Flat (non-aggregated) joins silently fan out events when the right side is not unique on the join key Open

bug

Difficulty 4/5 3-5 days Newbie friendliness 52/100

mmcdermott/MEDS_extract#184 · Opened Aug 4, 2026
Design: row-multiplying `explode` does not fit dftly's expression contract Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

mmcdermott/dftly#88 · 1 comment · Opened Jul 31, 2026
No list type: add a `split` node (and list accessors) so delimited cells can be expressed as List-valued expressions Open

Difficulty 5/5 Over a week Newbie friendliness 48/100

mmcdermott/dftly#87 · 3 comments · Opened Jul 31, 2026
Download-layer ergonomics: per-source fsspec storage options and optional pre-hash Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

mmcdermott/MEDS_extract#172 · Opened Jul 30, 2026
Support attaching metadata to literal codes (e.g. MEDS_BIRTH, MEDS_DEATH) Open

Difficulty 5/5 Over a week Newbie friendliness 30/100

mmcdermott/MEDS_extract#166 · Opened Jul 29, 2026
Design: extraction profiles + drift comparator — statistical regression testing for derived pipelines against proposed releases Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

mmcdermott/MEDS_extract#157 · Opened Jul 29, 2026
MTD_preprocess is not `python -m`-runnable, and shipped code spawns MEDS_transform-pipeline by bare PATH name Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

mmcdermott/meds-torch-data#115 · Opened Jul 27, 2026
Design (0.7): strings-first ingestion — read all CSV sources with infer_schema=False, type explicitly at use sites Open

Difficulty 5/5 Over a week Newbie friendliness 42/100

mmcdermott/MEDS_extract#143 · Opened Jul 23, 2026

Showing the newest 100

This page lists what was indexed most recently. Advanced filter has the whole inventory, narrowed by language, difficulty and how long a task takes.

Open advanced filter

mmcdermott

Showing the newest 100

Get new issues in your inbox