Skip to content

armgit / Soup

Public Imported from GitHub
0

Fine-tune LLMs from one YAML. Layer streaming trains an 8B model on a 4 GB laptop GPU.

docs(contributors): #2156 [skip ci]

Alpamysc069d2d2bccf
diff --git a/CONTRIBUTORS.md b/CONTRIBUTORS.md
index 30a60ede..2bf6cf68 100644
--- a/CONTRIBUTORS.md
+++ b/CONTRIBUTORS.md
@@ -943,6 +943,7 @@ Listed by first contribution. PR numbers link the work.
- Took the harness half of #1693 (the issue stays open for the measured table): `benchmarks/harness/stream_loss_chunk_vram.py` builds a chunked and an unchunked arm through the real `SFTTrainerWrapper.setup`, runs one streamed step each, and prints the pre-flight's own prediction, the free VRAM it compared against and the real peak per length. **They said up front they have no GPU, so it carries a CPU `--self-test`** that runs the same protocol on a tiny local Llama and checks 13 things (the realised length, the pre-flight's inputs and arithmetic, the hook on the chunked arm only, the verdicts and exit codes), and every figure in the new sizing text is labelled as the pre-flight's arithmetic, not a measurement. Review found the teardown of each arm unpinned and one repo scanner red on a `print()` assertion; the follow-up commit added a test that counts both teardown calls and the scanner's marker the same day. ([#1693](https://github.com/MakazhanAlpamys/Soup/issues/1693), [#2102](https://github.com/MakazhanAlpamys/Soup/pull/2102))
- Took #2047: GRPO could only score with a rule, so a task with no checkable answer had no scorer. They posted the plan first, took the maintainer's twelve corrections, then added `training.reward_fn: judge_rank`: one judge call sees all `num_generations` completions of a prompt, in a fresh seeded order, and the group scores `None` instead of 0.0 when the judge gives no usable answer. A required `grpo_judge_max_calls` is checked after the data loads and before the model does, and failed or all-equal groups are logged under `judge/`. **The part worth copying is the questions**: the PR lists the five choices it made without a ruling and the one mutation that survives, and the follow-up commit closed that survivor by deleting the dead flag instead of arguing for it. ([#2047](https://github.com/MakazhanAlpamys/Soup/issues/2047), [#2117](https://github.com/MakazhanAlpamys/Soup/pull/2117))
- Took item 3 of #797 (the issue stays open for the vision/audio k-bit test): `training.gradient_checkpointing: true` or a tier did nothing on `distill`, `classifier`, `asr` and `unlearn`, which never read it. They passed it to the HF arguments of the first three and, for unlearn's own loop, called `gradient_checkpointing_enable()` on the LoRA policy model, with tests that drive each real `setup()` on CPU and count the backward recompute, and a table of opposite-value mutations per site. 36 tests (26 fail on `main`), 10 of 10 mutations killed by named tests, and the adapter weights and first-step loss check out with checkpointing on. ([#797](https://github.com/MakazhanAlpamys/Soup/issues/797), [#2140](https://github.com/MakazhanAlpamys/Soup/pull/2140))
+ - Fixed the `expandable_segments` allocator hint never being applied to a streamed `soup train` on Linux: `detect_device()` started CUDA before the code that would set the variable ran, so every streamed CUDA run printed "not enabled" and the hint was lost. The hint is now set from `train.py` before device detection (streamed runs on Linux and WSL2 only, an exported value kept, an empty one counted as unset), and the reason line names the variable when another value is the cause. Verified on a WSL2 card by @abdulwaarith0 and with a recorder at `detect_device` on CPU; 12 of 12 source mutations killed ([#2149](https://github.com/MakazhanAlpamys/Soup/issues/2149), [#2156](https://github.com/MakazhanAlpamys/Soup/pull/2156))
- **Nariman Abdikarim** ([@Nari-Ab](https://github.com/Nari-Ab))
- Sanitized model and dataset paths across CLI display sinks and both short-name branches, with 17 portable regressions covering refusals, adapter loading, saved outputs and real Elo/leaderboard table consumers. Independent maintainer validation passed all 17 cases before and after five source undos that caused 1/1/1/3/4 failures; all eight required hosted checks passed ([#1864](https://github.com/MakazhanAlpamys/Soup/issues/1864), [#1868](https://github.com/MakazhanAlpamys/Soup/pull/1868))