RomanNum3ral
/
soup
mirror of https://github.com/razor-ai/soup.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

docs(readme): refresh Quant Menu section for v0.40.5 multi-trainer wiring 

The README's `## Quant Menu` body still said "v0.38.0 scope — wired into
the SFT trainer + transformers backend. Multi-trainer expansion is tracked
for v0.38.1." That stub-then-live note is now stale: v0.40.5 (#66) shipped
the multi-trainer wiring across all 11 non-SFT trainers + the PPO reward
model. Replace with the post-v0.40.5 status.

Docs-only hotfix — no soup_cli/ changes, no version bump per CLAUDE.md
"CI-only / docs-only hotfixes" policy.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Alpamys 2026-05-09 14:48:42 +05:00
parent 697cc8dad7
commit 0d406467a1
1 changed files with 7 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
README.md
View File

@ -543,10 +543,13 @@ for the full table.
`fp8` all carry their own scale; combining with `quantization_aware` (int8 QAT or
`'fp8'`) is rejected at config-load.
**v0.38.0 scope** — wired into the SFT trainer + transformers backend.
Multi-trainer expansion is tracked for v0.38.1 (mirrors v0.27.0 MII /
v0.37.0 multipack stub-then-live pattern). MLX backend gets a distinct error
message naming the actual reason.
**Multi-trainer support.** Quant Menu is wired across all 12 transformer-backend
trainers (SFT / DPO / GRPO / KTO / ORPO / SimPO / IPO / PPO / RewardModel /
Pretrain / Embedding / BCO). PPO's reward model also loads with the same Quant
Menu config as the policy when `tcfg` is passed in, so a GPTQ-policy + GPTQ-reward
run does not silently OOM in fp16. MLX backend is rejected with a distinct error
message; vision / audio modality is still SFT-only inline-BNB (multi-modal
Quant Menu wiring tracked as a follow-up).
## Multipack — FFD Bin-Packing Sampler