[codex] Add synchronous ES LoRA trainer#2447

Closed

Matthew-agi wants to merge 12 commits into

PrimeIntellect-ai:mainfrom

Matthew-agi:codex/synchronous-es-trainer

Commits on May 8, 2026

Add synchronous ES LoRA trainer

Matt
authored and
Matt
committed
Add concision Gemma ES environment

Matt
authored and
Matt
committed
Match prior one-sided ES update

Matt
authored and
Matt
committed
Add persistent vLLM LoRA slots for ES

Matt
authored and
Matt
committed

Commits on May 9, 2026

Add GSM8K ES configs

Matt
authored and
Matt
committed
Preserve GSM8K ES output directory on launch

Matt
authored and
Matt
committed
Add ES mean-policy eval support

Matt
authored and
Matt
committed
Add ES vLLM slot benchmark script

Matt
authored and
Matt
committed
Initialize dist in ES vLLM benchmark

Matt
authored and
Matt
committed
Add GSM8K vLLM sweep configs

Matt
authored and
Matt
committed
Tune GSM8K ES serving config

Matt
authored and
Matt
committed
Use Qwen3 base model for GSM8K configs

Matt
authored and
Matt
committed