Skip to main content

catgrad

You like category theory? You like tinygrad? You love catgrad! ❤️

catgrad is a bit different: instead of using autograd to train, you compile your model's reverse pass into static code. This means your training loop can run without needing a deep learning framework (not even catgrad!)

Here is a linear model in catgrad:

model = layers.linear(BATCH_TYPE, INPUT_TYPE, OUTPUT_TYPE)

catgrad can compile this model...

CompiledModel, _, _ = compile_model(model, layers.sgd(learning_rate), layers.mse)

... into static code like this...

class CompiledModel:
    backend: ArrayBackend

    def predict(self, x1, x0):
        x2 = x0 @ x1
        return [x2]

    def step(self, x0, x1, x9):
        x4, x10 = (x0, x0)
        x11, x12 = (x1, x1)
        x16 = self.backend.constant(0.0001, Dtype.float32)
        # ... snip ...
        x18 = x17 * x5
        x2 = x10 - x18
        return [x2]

... so you can train your model by just iterating step; no autograd needed:

for i in range(0, NUM_ITER):
    p = step(p, x, y)

Catgrad doesn't just compile to Python: I'm working on support for other targets like C++ (ggml), CUDA, FPGAs, and more!

Catgrad uses reverse derivatives and open hypergraphs to transform a model into its backwards pass. For details, see this paper.

Install

pip install catgrad

Examples

Train simple MLPs for the iris dataset:

./data/get-iris-data.sh
python3 -m examples.iris (linear|simple|dense|hidden)

Compilation Targets

Target backends we plan to support soon:

Metadata

Release files for catgrad 0.2.2

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

Source distribution (sdist)

Source distribution for catgrad 0.2.2
File Size Uploaded Provenance Actions
catgrad-0.2.2.tar.gz 18.4 kB Without provenance Details for catgrad-0.2.2.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for catgrad 0.2.2
File Interpreter ABI Platform Actions
catgrad-0.2.2-py3-none-any.whl Python 3 none any Details for catgrad-0.2.2-py3-none-any.whl

Total release size: 45.2 kB

Release files / catgrad-0.2.2.tar.gz

Download URL catgrad-0.2.2.tar.gz
Size 18.4 kB
Tags Source
SHA-256 checksum
How to use checksums
e4beb40aec2a3b1cfcb1a0ecb58113c580c0d02c058ade653429f61405d61a32
BLAKE2b-256 checksum
How to use checksums
0283cdba62d35a5784799b95f7042e20373a2203827445fb777b6cf0a7829008
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.1.0 CPython/3.13.1

Release files / catgrad-0.2.2-py3-none-any.whl

Download URL catgrad-0.2.2-py3-none-any.whl
Size 26.9 kB
Tags Python 3
SHA-256 checksum
How to use checksums
6730743f0f779256af6e5d0dcc9a0fd99dfb6c2b091aaa37d20cacf976ac7934
BLAKE2b-256 checksum
How to use checksums
262d9bc79e499653e8863e6c2e3a89d382e179002cb7bf19cb3a427a3bde83af
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.1.0 CPython/3.13.1

Security for catgrad 0.2.2

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