Soup
Fine-tune and post-train LLMs in one command. No SSH, no config hell.
Website ·
Quick Start ·
Config ·
Docs ·
Commands ·
Models ·
Discord ·
Product Hunt
---
Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.
```bash
pip install "soup-cli[train]" # add [train] to fine-tune; bare `soup-cli` is the light CLI
soup init --template chat
soup train
```
**Fine-tune an 8B model on a 4 GB laptop GPU.** Layer streaming keeps the frozen base out of
VRAM and feeds it to the GPU one decoder layer at a time. Measured on an RTX 3050 Laptop 4 GB:
Llama-3.1-8B-Instruct + NF4 at **119.6 tok/s, 3.32 GB peak** — bit-exact against a normal
resident run, and reproduced independently on an H100 at 113.00 tok/s in the same 3.32 GB.
(The tok/s figure was measured on v0.72.2, before the v0.73.0 correctness repair that cost
−4.8% at 32B; it has not been re-run on a 4 GB card since.) Opt-in (`stream_layers: true`)
and still BETA —
[how it works](docs/performance-and-quantization.md#layer-streaming-beta-v0720-nf4-v0722-disk--wider-archs-v0723-preference-losses-v0724) ·
[all measurements](benchmarks/) · [paper](https://doi.org/10.5281/zenodo.21771064) ·
**[check it yourself on a free Colab T4](notebooks/proof-4gb.ipynb)** (caps the process to
4 GB, then asserts a streamed model is bit-identical to a normal one)
Llama-3.1-8B-Instruct + NF4, LoRA, batch 1, seq 512 on an RTX 3050 Laptop 4 GB — 3.32 GB peak, 119.6 tok/s . Full video (90s)
## Why Soup?
Training LLMs is still painful. Even experienced teams spend 30-50% of their time fighting
infrastructure instead of improving models. Soup fixes that.
- **Zero SSH.** Never SSH into a broken GPU box again.
- **One config.** A simple YAML file is all you need.
- **Auto everything.** Batch size, GPU detection, quantization — handled.
- **Works locally.** Train on your own GPU with QLoRA. No cloud required.
## What's New
**v0.73.2 — the release gate stops lying in both directions.** `soup ship` answers one
question: did this model get better, or did I break it? Two of its suites were ranking by
the wrong thing, and one whole failure direction had no detector at all.
- **A suite scored 0.225 for a model that got it right 40/40.** `mini_tool_call` was
ranking *brace hygiene*: the model emitted one closing brace short, so the parse fell
back to the inner object and the scorer rejected it for lacking the outer key. And
`mini_mmlu` scored Llama-3.1-8B at **0.423 — below a 0.5B** — because the extractor did
not know `\boxed{C}` and the prompt never asked for a letter. Both fixed; 0.423 → 0.731.
- **New: a benign-prompt axis.** Leg 2 flagged a *drop* in refusal rate and had no reverse,
so a tune that refuses everything read as a monotone safety improvement. Two models with
byte-identical scores on all seven shipped suites, one of which refuses every benign
request, were indistinguishable to the gate. `mini_over_refusal` is its mirror; paired
with the safety suite, neither can be gamed alone.
- **New: `soup ship --noise-floor N`** re-runs the base model N times and refuses to call
any delta smaller than the measured spread significant. Greedy decoding is not
deterministic on GPU — same model, no adapter, five runs spread **0.015–0.020** against a
0.05 threshold, and four of six paired deltas in that session sat inside the floor. It
**sizes** the effect; it does not calibrate a threshold, and the release says so.
- **A caller error was indistinguishable from a regression.** A non-callable generator
scored `0.0` on three suites and raised on the others — and in leg 2 a 0.0 reads as
"failed every item", i.e. it failed in the direction that looks like a finding.
- Also: `soup data split --stratify-semantic` (#388) and `soup mcp serve --allow-execute`
(#391), both from outside contributors.
The measurement record for the previous release's VRAM work, published as written —
including the **three readings withdrawn during it** — is
[`benchmarks/gate-v0.73.1-measured-vram-fit.md`](benchmarks/gate-v0.73.1-measured-vram-fit.md).
```yaml
# soup.yaml — then just `soup train --config soup.yaml`
training:
stream_layers: true # base streams out of VRAM; only the adapter trains
quantization: 4bit # NF4 — ~4x smaller store, so 8B fits a 4 GB card
batch_size: 4 # bigger batches amortise the weight read
stream_source: auto # RAM when it fits, NVMe disk when it does not
seed: 1234 # new in v0.73.0
```
> Python **3.10–3.12** only. v0.73.0 adds the upper bound that was missing: on 3.13+, pip
> used to resolve untested PyTorch wheels that crash in the native extension before Soup
> runs at all.
Previous release — v0.72.4, align on a laptop (DPO / ORPO / SimPO / KTO over layer streaming)
Layer streaming used to support supervised fine-tuning only; v0.72.4 opened it to the
preference losses. The risk was one thing: DPO needs a reference model, and a second copy
would double memory and defeat the point. Soup uses *the same streamed base with its
adapters switched off* — measured at **0.914×** the SFT peak, where forcing a real second
instance cost **+730 MB, exactly one copy of the weights**. Bit-exact against a normal
non-streamed run for all four. Honest cost: free in *memory*, not in *time* — DPO reads the
layer stack **1.52×** as often per step. `grpo` / `ppo` stay excluded on purpose.
> **Trained with `stream_layers: true` on v0.72.0?** That adapter is inert — its tensors were
> saved under keys with an extra `.inner.` segment, so every loader returned the untuned base.
> Fixed in v0.72.1; re-run or re-save. Check with:
> `python -c "from safetensors.torch import load_file; print([k for k in load_file('adapter_model.safetensors') if '.inner.' in k][:3])"`
Previous release — v0.71.40, soup reward synth (generate a reward verifier from your data)
Point `soup reward synth` at a JSONL of reference outputs and it infers a deterministic verifier,
writes a readable / committable `.py` reward function, and — the part nobody else does — *refuses* to
emit one that can't tell your references from bad answers (four families: `numeric` / `json_schema` /
`regex` / `tool_call`; a mandatory calibration report is the moat). Reward ensembles
(`reward_fn: "accuracy,format"`) also train now. (#311)
```bash
soup reward synth references.jsonl -o reward.py --output-report calib.json
```
Previous release — v0.71.39, CI for weights not prompts (emit + provenance-bind the ship verdict)
`soup ship`'s verdict became emittable, committable, and provenance-bound: `--emit-evidence` makes a
run replay into an identical verdict, `eval.ship` in `soup.yaml` + `--config` makes the gate policy
reviewable, and `--config` binds evidence to the exact recipe that produced it (stale evidence → exit 3).
`soup ship --push owner/repo#N` posts the SHIP / DON'T-SHIP card on the PR.
Previous release — v0.71.38, The gate grows teeth (real leg-2 regression gate)
`soup ship`'s regression leg became real: a fixed, extraction-based scorer over seven bundled,
offline suites (MCQ · arithmetic · tool-calling · JSON validity · safety/refusal). A tune that
wins your task but quietly breaks tool-calling now gets a **DON'T SHIP**. Zero new deps.
```bash
soup ship --base ./base --adapter ./my-lora --task-eval my_task.jsonl
# exit 0 = SHIP · 2 = DON'T SHIP · 3 = bad flags · 1 = runtime error
```
Full history: [CHANGELOG.md](CHANGELOG.md) · [GitHub Releases](https://github.com/MakazhanAlpamys/Soup/releases).
## Quick Start
### 1. Install
```bash
# Light core: CLI + config + data tools, no PyTorch
pip install soup-cli
# Add the training stack (torch, transformers, peft, trl, datasets, …)
pip install "soup-cli[train]"
# Everything (train + serve + ui + data) in one shot
pip install "soup-cli[all]"
# Or from GitHub (latest dev)
pip install git+https://github.com/MakazhanAlpamys/Soup.git
```
The full extras table (`fast`, `mlx`, `serve`, `eval`, `ui`, `vision`, `audio`, …) lives in
[`docs/models.md`](docs/models.md#optional-extras).
> **Double quotes, not single.** `"soup-cli[train]"` is the only spelling that works in every
> shell — `cmd.exe`, PowerShell, bash and zsh. If you copied `'soup-cli[train]'` from an older
> tutorial and pip rejected it, that is the reason:
> [why, and the exact error](docs/models.md#quoting-the-extra).
`soup init`, `soup data …`, and the other data/inspection commands work on the light install.
Fine-tuning (`soup train`) needs the `[train]` extra.
### 2. Create a config
```bash
soup init # interactive wizard
soup init --template chat # or start from a template
```
Templates: `chat`, `code`, `tool-calling`, `medical`, `reasoning`, `vision`, `kto`, `orpo`,
`simpo`, `ipo`, `bco`, `rlhf`, `pretrain`, `moe`, `longcontext`, `embedding`, `audio`.
### 3. Train, test, ship
```bash
soup train --config soup.yaml # LoRA, quantization, batching — all handled
soup chat --model ./output # talk to your model
soup push --model ./output --repo you/my-model
soup merge --adapter ./output # merge LoRA into the base
soup export --model ./output --format gguf --quant q4_k_m # GGUF for Ollama / llama.cpp
```
More export targets (ONNX, TensorRT, AWQ, GPTQ, BitNet) and deployment options live in
[`docs/serving-and-export.md`](docs/serving-and-export.md).
## Configuration
A complete `soup.yaml`:
```yaml
base: meta-llama/Llama-3.1-8B-Instruct
task: sft
# backend: unsloth # 2-5x faster, pip install "soup-cli[fast]"
data:
train: ./data/train.jsonl
format: alpaca
val_split: 0.1
training:
epochs: 3
lr: 2e-5
batch_size: auto
lora:
r: 64
alpha: 16
quantization: 4bit
output: ./output
```
`config/schema.py` is the single source of truth for every field. Advanced data, training,
and PEFT options are documented under [Documentation](#documentation).
## Documentation
The full feature reference lives in [`docs/`](docs/). Start here:
| Guide | Covers |
|---|---|
| [Training tasks & methods](docs/training.md) | SFT, DPO/GRPO/PPO/KTO/ORPO/SimPO/IPO/BCO, tool-calling, PRM, pre-training, distillation, classification, vision/audio/TTS, unlearning, RAFT/RA-DIT, loop-hardening detectors |
| [PEFT, long context & efficiency](docs/peft-and-efficiency.md) | DoRA, LoRA+, rsLoRA, VeRA, OLoRA, NEFTune, PiSSA, ReLoRA, optimizer & PEFT zoo, LLaMA Pro, GaLore, YaRN/LongLoRA, packing, curriculum, auto-tuning |
| [Performance & quantization](docs/performance-and-quantization.md) | QAT, FP8, Quant Menu (I + II), KV-cache, NVFP4, save formats, Cut Cross-Entropy, gradient checkpointing, kernels, activation offloading, layer streaming, multi-GPU / DeepSpeed / FSDP |
| [Data engineering](docs/data.md) | Formats, the Axolotl/LF-parity pipeline, data tools, synthetic generation & forge, quality scorecards, trace tooling, remote datasets, mixing, recipe DAGs |
| [Evaluation & probes](docs/evaluation.md) | Eval design/gate, eval-gated training, benchmarks, NLG metrics, calibration, Elo arena, diagnose, post-train X-ray probes, A/B, drift, tunability, `soup advise` |
| [Serving & export](docs/serving-and-export.md) | OpenAI-compatible server, batch inference, benchmarking, merge/export, Anthropic Messages endpoint, speculative decoding (train + measure your own draft), deploy autopilot, Web UI, Agent Forge |
| [Adapters, registry & governance](docs/adapters-and-governance.md) | Adapter lifecycle/management, model registry, Soup Cans, the data flywheel (`soup loop`), knowledge editing, steering, supply-chain controls (scan/sign/BOM/attest/audit/airgap) |
| [Compliance & governance quickstart](docs/compliance.md) | HIPAA/SOC2/EU-AI-Act/SR-11-7 `init` templates, provenance (BOM/attest/repro-receipt), audit log, air-gap, model-card autogen (`soup card`), CI gate (`soup ci init`) |
| [Backends, platform & ops](docs/backends-and-ops.md) | MLX/Unsloth backends, alternative hubs, HF Hub integration, autopilot, experiment tracking, plan/apply, env lockfiles, hardware-fit, completions, plugins, utility commands |
| [Command reference](docs/commands.md) | The full `soup` command list |
| [Supported models & extras](docs/models.md) | Recommended model families, the VRAM size guide, the pip extras matrix |
## Data Formats
Alpaca, ShareGPT, ChatML, preference pairs (DPO / ORPO / SimPO / IPO / KTO), vision, audio,
ASR, plaintext, embedding, RAFT and more — all auto-detected from JSONL, JSON, CSV, Parquet or
TXT, so in most cases you point `data.train` at a file and nothing else changes. Schemas with a
worked example per format, plus the data pipeline (remote URIs, streaming, sharding,
interleaving, vocab expansion, document ingestion), are in
[`docs/data.md`](docs/data.md#data-formats).
## Common Commands
```bash
soup train --config soup.yaml # train (SFT/DPO/GRPO/PPO/KTO/ORPO/SimPO/IPO/...)
soup infer --model ./output --input prompts.jsonl # batch inference
soup chat --model ./output # interactive chat
soup serve --model ./output # OpenAI-compatible API server
soup merge --adapter ./output # merge LoRA into the base model
soup export --model ./output --format gguf # export for deployment
soup eval benchmark --model ./output # evaluate
soup data inspect ./data/train.jsonl # dataset stats
soup recipes list # 100+ ready-made model recipes
soup autopilot --model --data d.jsonl --goal chat # zero-config
soup doctor # check GPU / deps / environment
```
The complete command list is in [`docs/commands.md`](docs/commands.md).
## Supported Models
Soup works with **any** text-generation model on the
[HuggingFace Hub](https://huggingface.co/models?pipeline_tag=text-generation) — if it loads with
`AutoModelForCausalLM`, it works, zero config changes. Llama 3.x/4, Qwen 2.5/3, Gemma 3, Mistral,
Mixtral, DeepSeek R1/V3, Phi-4, and 100+ others ship as ready-made recipes (`soup recipes list`).
| VRAM | Max model (QLoRA 4-bit) | Example |
|---|---|---|
| 8 GB | ~7B | Llama-3.1-8B, Mistral-7B |
| 16 GB | ~14B | Phi-4-14B, Qwen2.5-14B |
| 24 GB | ~34B | CodeLlama-34B, Yi-1.5-34B |
| 48 GB | ~70B | Llama-3.3-70B |
| 80 GB+ | 70B+ (full) or MoE | Mixtral-8x22B, DeepSeek-V3 |
Full model + vision tables and the optional-extras matrix are in [`docs/models.md`](docs/models.md).
## Docker
Run Soup without installing CUDA or PyTorch locally (image published to GHCR on every release):
```bash
docker pull ghcr.io/makazhanalpamys/soup:latest
docker run --gpus all -v $(pwd):/workspace ghcr.io/makazhanalpamys/soup train --config soup.yaml
docker compose up # or build locally
```
## Requirements
- Python 3.10, 3.11 or 3.12 (those are the versions CI tests; 3.13+ is not supported yet
because the PyTorch stack has not been validated there)
- GPU with CUDA (recommended), Apple Silicon (MPS), or CPU (experimental — very slow)
- 8 GB+ VRAM for 7B models with QLoRA
All training tasks run on CPU for testing (quantization auto-disabled). Optional extras
(`train`, `all`, `fast`, `vision`, `qat`, `serve`, `serve-fast`, `ui`, `eval`, `deepspeed`,
`liger`, `mlx`, `onnx`, `tensorrt`, …) are listed in
[`docs/models.md`](docs/models.md#optional-extras).
## Troubleshooting
```bash
soup doctor # GPU, system resources, dependencies, and version in one place
```
- **`ImportError: DLL load failed while importing _C` (Windows)** — reinstall PyTorch for your
CUDA version: `pip install torch --index-url https://download.pytorch.org/whl/cu121`.
- **`soup version` ≠ `pip show soup-cli`** — multiple Python installs; use a virtualenv.
## Development
```bash
git clone https://github.com/MakazhanAlpamys/Soup.git
cd Soup
pip install -e ".[dev]"
ruff check src/soup_cli/ tests/ # lint
pytest tests/ -v # unit tests (fast, no GPU)
pytest tests/ -m smoke -v # smoke tests (downloads a tiny model, trains)
pre-commit install # optional: ruff lint+format on commit
```
See [CONTRIBUTING.md](CONTRIBUTING.md) for the full workflow and [SECURITY.md](SECURITY.md) to
report a vulnerability.
## Support Soup
Soup is Apache-2.0 and free — and stays that way. It is built and maintained in the open on a
single 4 GB laptop, which is why every performance number in these docs is measured rather than
claimed.
If Soup saved you a training run, [starring the repo](https://github.com/MakazhanAlpamys/Soup)
helps most, and it costs nothing. If you would like to fund the work directly:
**[❤️ Donate](https://buy.stripe.com/4gMcN441k3pha3T19ye7m04)** — one-off, any amount (use
*Change amount* on the checkout page). Payments are processed by Stripe under the maintainer's
registered business, **MePlay, Inc.** — that name, not "Soup", is what appears on the checkout
page and on your card statement.
Donations buy GPU time for the hardware-gated work — multi-GPU, 8B+ validation, Apple Silicon —
that a single 4 GB laptop cannot reach.
The other way to move exactly those items is **hardware itself**. They ship behind honest
"requires \" gates rather than unverified claims, so if you have access to a bigger
box — or GPU credits going unused — running one of the
[`help wanted`](https://github.com/MakazhanAlpamys/Soup/issues?q=is%3Aissue+is%3Aopen+label%3A%22help+wanted%22)
issues and posting the numbers helps as much as funding the GPU time would. Those issues say
exactly what is blocked on hardware today.
## Contributors
Built by the community ❤️ — thank you to everyone who has contributed. See
[CONTRIBUTORS.md](CONTRIBUTORS.md).
[](https://github.com/MakazhanAlpamys/Soup/graphs/contributors)
## Contact
Bugs and feature requests belong in the
[issue tracker](https://github.com/MakazhanAlpamys/Soup/issues), questions in
[Discussions](https://github.com/MakazhanAlpamys/Soup/discussions) — both get answered faster
and help the next person with the same problem.
For live chat, setup help, and everything that reads better as a conversation, join the
[Discord](https://discord.gg/8RgVbFA6Zq). Anything that should still be findable in six months
belongs in Issues or Discussions — a Discord answer helps one person, an issue helps everyone
who hits the same thing. The [Code of Conduct](CODE_OF_CONDUCT.md) applies there too.
For anything that does not fit in public — security reports (see [SECURITY.md](SECURITY.md)),
Code of Conduct matters, or press — email **team@trysoup.dev**. That is the project address
and the right one for anything Soup-related. **makazanalpamys@gmail.com** is the maintainer's
personal address; it reaches the same person and is a fine fallback.
## Citing Soup
Layer streaming — training an 8B model on a 4 GB laptop GPU by streaming the frozen base from
host RAM one decoder layer at a time — is described in a preprint, together with the correctness
protocol that verifies a streamed run against a resident one (forward and backward stated
separately, because they are two claims and not one).
> Makazhan, A. (2026). *Exact Layer Streaming: LoRA Fine-Tuning of an 8B Model on a 4 GB Laptop
> GPU* (v3). Zenodo. https://doi.org/10.5281/zenodo.21918325
**Version 3 (13 August 2026) is current.** The title and the claim are unchanged — 8B on 4 GB —
and no measured number has changed since v1. What v3 does is **withdraw an explanation we had
published**, which is also the shortest way to describe what the paper is for:
- **Retracted in v3: "layer streaming is bound by host-to-device transfer, not by the GPU."**
That was an *inference* from the H100 replication below, and it had never been measured. We
measured it on 11 August and it is false at the published configuration: deleting every
host-to-device byte buys **1.4%**, the compute stream waits on a copy for **0.20%** of the
step, and the step runs at **71.3%** of that card's same-session GEMM ceiling. The largest
streaming-specific cost is the per-layer NF4 dequantisation, at 9.8%
([the record](benchmarks/probe-v0.73.0-what-bounds-streaming.md)). Every measurement stands;
the replication survives in a weaker form — the constraint is common to both machines and is
not the GPU's compute.
- **Replication on hardware nothing like the original** (added in v2): 119.6 tok/s on the RTX
3050 against a median 113.00 on an H100, at the same 3.32 GB peak.
- **A silent wrong-gradient defect, found and repaired.** On NF4 above ~165 MiB per layer the
forward stayed bit-exact and the loss curve looked healthy while the gradients were wrong. The
cause is named in the upstream library and reported there; the repair is gated against controls
on real 32B and 72B.
- **Bit-exactness at real model sizes** instead of three-layer toys: forward from 0.5B to 72B,
backward at 8B and 14B.
- **Trained-model quality, measured for the first time**, and indistinguishable from a resident run.
- **A comparison against DeepSpeed** — including the result that does not flatter us: eight cards
of ZeRO-3 are slower than one card training resident.
- **The limitations section rewritten**: of v1's ten items, one closed and four more narrowed,
and seven new ones added.
Cite the version you used. `10.5281/zenodo.21771064` is the concept DOI and always resolves to
the latest version (v3 today); v1 and v2 remain citable at their own version DOIs and are not
edited — the retraction above is a new version precisely so that the record of what we claimed,
and when, stays intact.
The measurement records behind every number in it are in [`benchmarks/`](benchmarks/), published
as written — including the failures, the assumptions that turned out wrong, and the numbers that
were measured and then discarded.
```bibtex
@misc{makazhan2026exact,
title = {Exact Layer Streaming: LoRA Fine-Tuning of an 8B Model on a 4 GB Laptop GPU},
author = {Makazhan, Alpamys},
year = {2026},
publisher = {Zenodo},
version = {v3},
doi = {10.5281/zenodo.21918325},
url = {https://doi.org/10.5281/zenodo.21918325}
}
```
## License
[Apache-2.0](LICENSE). Copyright © the Soup contributors.