Repository navigation

Add a multi-turn GRPO example with tool calling, and port OPSD to the per-sample rollout API - #1023

Open

anyadontfly wants to merge 5 commits into

deepspeedai:masterfrom

anyadontfly:grpo

Open

Add a multi-turn GRPO example with tool calling, and port OPSD to the per-sample rollout API#1023
anyadontfly wants to merge 5 commits into
deepspeedai:masterfrom
anyadontfly:grpo

Commits on Nov 7, 2026
- Add multi-turn GRPO example with tool calling
  anyadontfly
  committed
  11445db
Commits on Nov 8, 2026
- Train Qwen3-4B-Instruct-2507 on one GPU and fix the interpreter's checks
  anyadontfly
  committed
  21a1ccf
- Add a notebook-style option to the code interpreter
  anyadontfly
  committed
  51c87a2
Commits on Nov 9, 2026
- Port the OPSD example to the per-sample rollout API
  anyadontfly
  committed
  47b4f2e
- Leave grouped-query decode attention to the framework
  anyadontfly
  committed
  5c1bc1a