pullrequests Search Results · language:Dune language:Python language:JavaScript language:HTML language:CSS language:PHP
Filter by
196M results
Summary
Refreshes CLAUDE.md so it matches the current state of the repository. The previous version referenced v1.7–v1.9 but the
repo is now at v1.9.2, and several skills and the development workflow ...
Bumps undici from 8.1.0 to 8.4.0. details summary Release notes /summary p em Sourced from a href=
https://github.com/nodejs/undici/releases undici s releases /a . /em /p blockquote h2 v8.4.0 /h2 h2 What ...
dependencies
javascript
Resumen
Se agregan métricas estadísticas avanzadas para complementar el análisis de movimientos financieros.
Cambios realizados
- Implementación de la función mediana().
- Implementación de la función ...
Bumps @types/node from 22.19.19 to 25.9.2. details summary Commits /summary ul li See full diff in a href=
https://github.com/DefinitelyTyped/DefinitelyTyped/commits/HEAD/types/node compare view /a /li ...
dependencies
javascript
Summary
- Introduce v3/js/main.js as the V3 module entrypoint.
- Switch v3/index.html from multiple script defer tags to a single type= module script.
- Convert low-dependency files to ES module ...
Problem
The FPI service was crashing on startup because db/mysql.js hardcoded localhost as the database host and used
non-standard DB_HOST/DB_USER/etc. variable names, so it never picked up the Railway ...
tested to work on my new pellet gril. I added the grill to my wifi using the tuya app, then used the Tuya cloud API
access to verify all the device items were correct for this specific device model.
Replaces autofix-ci with amend+forcepush to keep a clean single-commit history on PR branches.
Enforces:
- PR title ≤60 characters
- PR body lines ≤72 characters
- Exactly one commit per PR
Summary
- Exact score (⭐ match) now uses fractional points: +3 exact, +2 under ±1, +1.5 over ±1, +0.5 off by 2, 0 off by 3+
- Automatic tiebreaker: under-predictor beats over → closest on losing team ...
reproducing: PR
- ReLU^2
- zero init proj layer
- qk norm changed head dim to 128 for qk norm to be effective
- also removed grad clipping
At 1B tokens, val_loss is 3.46 vs baseline 3.70