MakazhanAlpamys/Soup

🌍 English | Türkçe | العربية | 日本語 | 中文

Fine-tune and post-train LLMs in one command. No SSH, no config hell.

Website · Quick Start · Web UI · Config · Docs · Commands · Models · Discord · Telegram · Product Hunt

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

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 at 952 MHz, 3.32 GB peak (pre-repair) — bit-exact against a normal resident run, and reproduced independently on an H100 at 113.00 tok/s in the same 3.32 GB. (Both figures were measured on v0.72.2, before the v0.73.0 correctness repair that cost −4.8% at 32B. The post-repair run, on another RTX 3050 Laptop 4 GB (@umran666, #1103), reads 208.6 tok/s at 1935–1957 MHz in a 2.40 GB peak. That is 59% of its same-session GEMM ceiling against 68%: about 14% slower per unit clock, so the raw 1.74x gain is the clock, not a speedup. Large-layer streaming also changed the memory pattern between the two rows, so even the fraction is not like-for-like — the record.) Opt-in (stream_layers: true) and still BETA — how it works · all measurements · paper · check it yourself on a free Colab T4 (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 at 952 MHz (pre-repair, v0.72.2; post-#331: 208.6 tok/s at 1935 MHz in 2.40 GB, not directly comparable, @umran666, #1103). Full video (90s)}

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.

v0.76.0 — settings Soup accepted and never applied now stop the load. Options that were validated and documented, and read by nothing on the path they claimed to configure, are now refused at load, naming the setting and what to change; the rest warn and name the release that will refuse them. A config that loaded before can stop loading: run soup doctor --config soup.yaml first.

• Breaking: a setting that never took effect is refused at load. Among them training.quantization_aware: true (it never ran quantization-aware training), training.use_longlora, training.ebft_variant, training.gdpo_variant, training.rope_scaling_type where no setup reads it, freeze_layers / freeze_ratio outside the text sft / tts path on the transformers backend, auto_mixed_precision outside sft / tts, and backend: unsloth on seven further tasks that have no unsloth setup. Each message names the field and what to change. lr_groups, early_stop_patience, forgetting_eval_steps and other unread knobs warn at load now and are refused from v0.77.
• Validation loss is now evaluated. data.val_split held rows out of training on every run, and no trainer on the transformers backend evaluated them. Every task that gets a split now evaluates it at the end of each epoch, or every N steps with training.eval_steps, and records the val loss; runs take a few percent longer. Breaking for grpo, ppo and online_dpo: by default they no longer hold rows out (data.val_split is ignored with a note), so every row trains.
• Breaking: every non-standard grpo_variant applies the grpo_beta KL penalty. gspo, dapo, dr_grpo, bnpo, two_sided and rft validated grpo_beta and dropped it, so existing variant configs will not reproduce prior runs. grpo_beta: 0 now runs KL-free and skips the reference forward pass.
• Layer streaming reads its disk tier on a background thread, through direct I/O, and can stripe it across NVMe drives. On a 36 GB 70B-shaped NF4 store larger than RAM (one RTX 5070 Laptop 8 GB box), a cold step went from 110.5 s on the synchronous read to 16.2-17.3 s, and from 17.4 s to 9.97 s across two drives with SOUP_LAYER_STREAM_STRIPE_DIRS (a sustained-load probe: two rounds, 1.28x and 1.41x, no formal verdict; record). A store that fits RAM is still faster on the RAM tier, which stream_source: auto picks. Opt-in (stream_layers: true) and still BETA; records: direct I/O, striping.
• New commands and options. soup adapters check scans a LoRA adapter for silent failure; soup bench train trains a fixed number of steps and exits 1 when the model was not training (it covers task: sft on the transformers backend; soup bench is now a group, and soup bench infer is the old inference benchmark); soup data clean sanitizes and repairs a dataset without touching the input file; soup recipes verify checks on the meta device, with no weights downloaded, that every shipped recipe can attach a LoRA adapter; soup rewind names the dataset rows behind a loss spike. soup infer gains opt-in --batch-size and an experimental --cuda-graphs (PyTorch 2.14 or newer; on one RTX 5070 Laptop, unpinned, Qwen2.5-1.5B in FP16, the slowest run with it decoded 2.9x faster than the fastest without). soup ab runs a two-sided, variance-robust test, and soup adapters sign --backend sigstore and soup attest emit --sign sigstore sign with Sigstore keyless OIDC (pip install "soup-cli[sigstore]").
• Deprecated: soup export --format awq and --format gptq will be removed in v0.77; both upstream projects are archived and neither extra installs next to [train].

Python 3.10–3.12 only. On 3.13+, pip used to resolve untested PyTorch wheels that crash in the native extension before Soup runs at all.

Older highlights live on the GitHub Releases page.

Soup is a command-line application, so the cleanest install gives it its own environment and puts soup on your PATH:

# Light core: CLI + config + data tools, no PyTorch
pipx install soup-cli
uv tool install soup-cli          # same idea, if you already use uv

# Add the training stack (torch, transformers, peft, trl, datasets, …)
pipx install "soup-cli[train]"

# Everything (train + serve + ui + data) in one shot
pipx install "soup-cli[all]"

# Or from GitHub (latest dev)
pipx install "git+https://github.com/MakazhanAlpamys/Soup.git"

Already inside a virtualenv, a Colab notebook, or a Docker image? Use pip directly, with the same names and extras:

pip install soup-cli
pip install "soup-cli[train]"
pip install "soup-cli[all]"
pip install git+https://github.com/MakazhanAlpamys/Soup.git

Use pip rather than pipx if you also want to import soup_cli from your own code, since pipx deliberately isolates the application from everything else.

The full extras table (fast, mlx, serve, eval, ui, vision, audio, …) lives in docs/models.md.

error: externally-managed-environment? That is PEP 668, not a Soup problem. Debian 12, Ubuntu 23.04 and later stop pip from writing into the system Python, because apt manages those files too. pipx and uv tool sidestep it by giving Soup its own environment, which is why they are listed first above. python3 -m venv .venv && source .venv/bin/activate then plain pip works just as well.

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.

soup init, soup data …, and the other data/inspection commands work on the light install. Fine-tuning (soup train) needs the [train] extra.

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.

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.

Prefer a browser? soup ui serves a local dashboard for experiments, training setup, live metrics, dataset exploration and model chat.

pip install "soup-cli[ui]"
soup ui
# Opens http://127.0.0.1:7860/?token=<token> and prints the same URL

Web UI documentation

A complete soup.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.

Unknown config keys are rejected since v0.75. A key no model declares — a typo like quantizaton, or a field that only exists on a newer Soup — used to validate clean and be discarded, so the run proceeded with the setting simply not applied. v0.74 reported it at load with the field you probably meant; from v0.75 the same config fails to load, so fix or remove the key rather than relying on it being ignored. See Unknown config keys.

The full feature reference lives in docs/. Start here:

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, with native Sesame CSM audio training documented separately.

soup train  --config soup.yaml        # train (SFT/DPO/GRPO/PPO/KTO/ORPO/SimPO/IPO/...)
soup infer  --model ./output --input prompts.jsonl --output results.jsonl   # batch inference
soup chat   --model ./output          # interactive chat
soup serve  --model ./output          # OpenAI-compatible API server
soup ui                               # local browser dashboard
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 <id> --data d.jsonl --goal chat  # zero-config
soup doctor                           # check GPU / deps / environment

The complete command list is in docs/commands.md.

Soup works with any text-generation model on the HuggingFace Hub — 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).

Full model + vision tables and the optional-extras matrix are in docs/models.md.

Run Soup without installing CUDA or PyTorch locally (image published to GHCR on every release):

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

• 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.

soup doctor    # GPU, system resources, dependencies, and version in one place

CUDA wheels, version mismatches: docs/backends-and-ops.md.

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)
pytest tests/ -m gpu --no-cov -v   # GPU tests (need a CUDA card; report results, see CONTRIBUTING.md)

pre-commit install                 # optional: ruff lint+format on commit

See CONTRIBUTING.md for the full workflow and SECURITY.md to report a vulnerability. Telemetry is strictly opt-in (SOUP_TELEMETRY=1, default off; see Privacy Policy).

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 helps most, and it costs nothing. If you would like to fund the work directly:

❤️ Donate — 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 <hardware>" 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 issues and posting the numbers helps as much as funding the GPU time would. Those issues say exactly what is blocked on hardware today.

Built by the community ❤️ — thank you to everyone who has contributed. See CONTRIBUTORS.md.

Bugs and feature requests belong in the issue tracker, questions in 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 or the Telegram community. 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 applies there too.

For anything that does not fit in public — security reports (see 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.

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). 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 (at 952 MHz) against a median 113.00 on an H100, at the same 3.32 GB peak. Both pre-date the #331 repair and are the figures the paper reports. The post-#331 row (@umran666, #1103) is 208.6 tok/s at 1935–1957 MHz in a 2.40 GB peak, 59% against 68% of the same-session GEMM ceiling: the raw 1.74x gain is the clock, not a speedup, and the memory pattern changed between the rows, so even the fraction is not like-for-like (the record).
• 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/, published as written — including the failures, the assumptions that turned out wrong, and the numbers that were measured and then discarded.

@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}
}

Apache-2.0. Copyright 2026 Makazhan Alpamys and the Soup contributors. If you redistribute Soup or a modified copy of it, sections 4 and 6 of the license apply to you: keep LICENSE and NOTICE with every copy, mark the files you changed, and do not use the Soup name beyond saying where the work comes from.