Skip to main content

Tinker Cookbook

We provide two libraries for the broader community to customize their language models: tinker and tinker-cookbook.

  • tinker is a training SDK for researchers and developers to fine-tune language models. You send API requests to us and we handle the complexities of distributed training.
  • tinker-cookbook includes realistic examples of fine-tuning language models. It builds on the Tinker API and provides common abstractions to fine-tune language models.

Installation

  1. Sign up for Tinker through the waitlist.
  2. Once you have access, create an API key from the console and export it as environment variable TINKER_API_KEY.
  3. Install tinker python client via pip install tinker
  4. We recommend installing tinker-cookbook in a virtual env either with conda or uv. For running most examples, you can install via pip install -e ..

Tinker

Refer to the docs to start from basics. Here we introduce a few Tinker primitives - the basic components to fine-tune LLMs:

import tinker
service_client = tinker.ServiceClient()
training_client = service_client.create_lora_training_client(
  base_model="meta-llama/Llama-3.2-1B", rank=32,
)
training_client.forward_backward(...)
training_client.optim_step(...)
training_client.save_state(...)
training_client.load_state(...)

sampling_client = training_client.save_weights_and_get_sampling_client(name="my_model")
sampling_client.sample(...)

See tinker_cookbook/recipes/sl_loop.py and tinker_cookbook/recipes/rl_loop.py for minimal examples of using these primitives to fine-tune LLMs.

To download the weights of any model:

rest_client = service_client.create_rest_client()
future = rest_client.get_checkpoint_archive_url_from_tinker_path(sampling_client.model_path)
with open(f"model-checkpoint.tar.gz", "wb") as f:
    f.write(future.result())

Tinker Cookbook

Besides these primitives, we also offer Tinker Cookbook (a.k.a. this repo), a library of a wide range of abstractions to help you customize training environments. tinker_cookbook/recipes/sl_basic.py and tinker_cookbook/recipes/rl_basic.py contain minimal examples to configure supervised learning and reinforcement learning.

We also include a wide range of more sophisticated examples in the tinker_cookbook/recipes/ folder:

  1. Chat supervised learning: supervised fine-tuning on conversational datasets like Tulu3.
  2. Math reasoning: improve LLM reasoning capability by rewarding it for answering math questions correctly.
  3. Preference learning: showcase a three-stage RLHF pipeline: 1) supervised fine-tuning, 2) learning a reward model, 3) RL against the reward model.
  4. Tool use: train LLMs to better use retrieval tools to answer questions more accurately.
  5. Prompt distillation: internalize long and complex instructions into LLMs.
  6. Multi-Agent: optimize LLMs to play against another LLM or themselves.

These examples are located in each subfolder, and their README.md files will walk you through the key implementation details, the commands to run them, and the expected performance.

Import our utilities

Tinker cookbook includes several utilities. Here's a quick overview:

  • renderers converts tokens from/to structured chat message objects
  • hyperparam_utils helps calculate hyperparameters suitable for LoRAs
  • evaluation provides abstractions for evaluating Tinker models and inspect_evaluation shows how to integrate with InspectAI to make evaluating on standard benchmarks easy.

Contributing

This project is built in the spirit of open science and collaborative development. We believe that the best tools emerge through community involvement and shared learning.

We welcome PR contributions after our private beta is over. If you have any feedback, please email us at tinker@thinkingmachines.ai.

Citation

If you use Tinker for your research, please cite it as:

Thinking Machines Lab, 2025. Tinker. https://thinkingmachines.ai/tinker/.

Or use this BibTeX citation:

@misc{tml2025tinker,
  author = {Thinking Machines Lab},
  title = {Tinker},
  year = {2025},
  url = {https://thinkingmachines.ai/tinker/},
}

Metadata

Release files for tinker-cookbook 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for tinker-cookbook 0.1.0
File Size Uploaded Provenance Actions
tinker_cookbook-0.1.0.tar.gz 669.0 kB Without provenance Details for tinker_cookbook-0.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for tinker-cookbook 0.1.0
File Interpreter ABI Platform Actions
tinker_cookbook-0.1.0-py3-none-any.whl Python 3 none any Details for tinker_cookbook-0.1.0-py3-none-any.whl

Total release size: 1.0 MB

Release files / tinker_cookbook-0.1.0.tar.gz

Download URL tinker_cookbook-0.1.0.tar.gz
Size 669.0 kB
Tags Source
SHA-256 checksum
How to use checksums
0e60f934e73b7de64b6f54b0dcaec0d7fa4f0025525ddb5dfb47e1605fd3979f
BLAKE2b-256 checksum
How to use checksums
3bca99cf38c9e2b5abaadeeb522d1018aa37de76920bc3359eab07cf614a68e7
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.6.9

Release files / tinker_cookbook-0.1.0-py3-none-any.whl

Download URL tinker_cookbook-0.1.0-py3-none-any.whl
Size 379.3 kB
Tags Python 3
SHA-256 checksum
How to use checksums
47b3f71e831758452be7950a132454314cdf3b688eba262403dc84365cf9b417
BLAKE2b-256 checksum
How to use checksums
2be271220ef68b2c7ddec23701ca3ff9312ec4b9f6e5603b5dfc5c1b7ab53814
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.6.9

Security for tinker-cookbook 0.1.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page