PyPI · #3375 most downloaded on PyPI

cog

0.23.01.9M downloads/moreplicate/cog ↗New releases in your daily digest, changelog included.

Containers for machine learning

Last release 9 days ago
22 Sep 2026
Release timing varies
gaps range from 8 days to 2 months
Nearly every release is documented
notes for 59 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
5 years old
184 releases · first in 2021

Release timeline

184 releases since 2021

202120222023202420252026

20212026

One column per quarter.

Releases

0.14.114 Mar 2025
9a05507f89d40e069ca71b63e3b08e918a6696f0 Make progress bar more pretty
Changelog

9a05507f89d40e069ca71b63e3b08e918a6696f0 Make progress bar more pretty (#2198)

955912d1fe2e51e34d4d9136fb9aee94c1e62f0b Support pulling monobase in macOS (#2197)
Open source →
0.14.013 Mar 2025Breaking
This is now unsupported and will receive no further updates. There are some breaking changes in the API that will be introduced with the release of th…
Support for concurrent predictions

This release introduces support for concurrent processing of predictions through the use of an async predict function.

To enable the feature add the new concurrency.max entry to your cog.yaml file:

concurrency: max: 32

And update your predictor to use the async def predict syntax:

class Predictor(BasePredictor): async def setup(self) -> None: print("async setup is also supported...") async def predict(self) -> str: print("async predict"); return "hello world";

Cog will now process up to 32 predictions simultaneously, once at capacity subsequent predictions will return a 409 HTTP response.

Iterators

If your model is currently using Iterator or ConcatenateIterator it will need to be updated to use AsyncIterator or AsyncConcatenateIterator respectively.

from cog import AsyncConcatenateIterator, BasePredictor class Predict(BasePredictor): async def predict(self) -> AsyncConcatenateIterator[str]: for fruit in ["apple", "banana", "orange"]: yield fruit

Migrating from 0.10.0a

An earlier fork of cog with concurrency support was published under the 0.10.0a release channel. This is now unsupported and will receive no further updates. There are some breaking changes in the API that will be introduced with the release of the 0.14.0 beta. This alpha release is backwards compatible and you will see deprecation warnings when calling the deprecated functions.

emit_metric(name, value) - this has been replaced by current_scope().record_metric(name, value)

[!NOTE] Note that the use of current_scope is still experimental and will output warnings to the console. To suppress these you can ignore the ExperimentalFeatureWarning:

import warnings from cog import ExperimentalFeatureWarning warnings.filterwarnings("ignore", category=ExperimentalFeatureWarning)

Known limitations

An async setup method cannot be used without an async predict method. Supported combinations are: sync setup/sync predict, async setup/async predict and sync setup/async predict.

File uploads will block the event loop. If your model outputs File or Path types these will currently block the event loop. This may be an issue for large file outputs and will be fixed in a future release.

Changelog

efad169ea7bc6113cd7d19c3a429ae4f6690b9ac Add a simple explanation for the standard predictor (#2187)

7b97da533e9351e1e4a363c545e30662b145f345 Add fast pusher for fast builds (#2114)

8a6fa4cd9efb57a2793251fb50520a07c38ed4fc Add multi progress bar to uploads (#2134)

ecbe1783962a396c784a6fd91d0da73f662a29e0 Add upload and verification to the fast pusher (#2128)

663c37503cdeddd34a53ca23b83ae3621cd9684d Build framework for file challenge (#2175)

f99ef7354bbbb7f38544b7f7c56879b1182654d5 Bump github.com/golangci/golangci-lint from 1.62.2 to 1.64.2

6f55fdd8e260c8fc7e9cdabd8f5cf308eb76c626 Call the web when layers have been pushed (#2139)

18a3e90fe4fb4378ddab5d85fff5eb62c3f6f323 Capture standard output when loading the predictor

85658ae1314e558c05aecf15a01b0f818ee6cdae Change ffmpeg int test to 3.12 (#2131)

9fe30d71e3025fffd063307233f4e3ad4a422cb0 Check version formats in fast_generator (#2129)

c95774f18b8df6f6a7b841be1900153ca04a57ad Create multipart uploader for fast push (#2135)

59545ccf89db80f9d2cb059191ce4233c83ff2a4 Enforce usage of r8.im image name (#2146)

b3713a688fbe95581c8ad5cb87577209cc3e8e03 Fail python_packages on --x-fast (#2162)

552a0cc5f8660941103abd64dfba50336f3d8852 Fast Push Fixes (#2133)

3f0d7c9cb101e63c1d6f6e77d2c16ad42485b386 Fix fast tarball tmp dir (#2194)

0e36b612949619585edbae6abe86982846adc5fd Fix file too long on tmp file create (#2132)

1247b483425ee73ef9dcaa2ba7e861934cb5dbf6 Fix monobeam client not authenticating properly (#2148)

018aca50f9d9d43ff784f1ea239c7cede72dc7c0 Fix path for weights in symlink (#2172)

c245552ad1e1904681f77a9dd50f57cfc8b44557 Freeze user layer only for fast builds (#2165)

e7e042cfba31ee2701f56c05e98539afe8d61f01 Handle URL Path in cog predict

61d9b8d6b8426f77fad231f735f4ed29309097e3 Handle weights moved to subdir (#2191)

e12f2374827d64a3e3b90c96436bddf8963eda58 Include sha256: prefix in generated runtime config files (#2153)

80d2a273050de720afdf03c972f361c62893771c Move fast push test to cog-runtime (#2181)

3ab642a3e9b634a235ba2d95fb36e0305818222a Propagate R8_COG_VERSION on fast push (#2166)

e0e8b6791e368a009c024cafb471670c1756f189 Pull monobase:latest before building fast model

425715c0b71009ce1158de02e7639da67cfd5813 Remove COG_EAGER_IMPORTS from fast path

c417cec53fc4471fbdbe3756bb736029cc744ca3 Remove COG_PYENV_PATH from fast path

2a0763ac6342513fc40c174e63b0dc37ff562a77 Remove deprecated interfaces Predictor.log and emit_metric

94bd7be1a063db70383cd9f420750fed54ad652c Require a test to be added for fast push (#2149)

a0859095f600b691c667b7238e3d09052115bfae Revert "Handle URL Path in cog predict"

0c6c0bed2457f49e684d842f9bc82a16f9b5494c Send push timings to the server (#2152)

e27d743a53de166d59c6deb510a21a0d96b7d933 Separate tmp mounts for fast layers (#2182)

bc4b95e070b271e06d796d335986791908271fd0 Support Python requirements package[extra,...]==version (#2160)

a27eb98f0a748b896db9258b174b8869b12bc81b Tweak fast build cache and file exposure (#2188)

cca8874bdf5ca5c6018422d469461ca343ee161f Update API endpoint in accordance with new name (#2151)

b49e4057a7698a8c368d1c2eb3425434aa418ea4 Update CDN URL (#2161)

06016884334c9e13fb0d1f43abe5cd2f9dc2e069 Update ruff config so lint works again (#2173)

8cfd9e15e33a5bc2d1d4ab76e7c1ce9fc81edb84 Update training.md

aed8fe0c096077a8ec5a17029a2bde580590e402 Use coglet in fast_generator (#2130)

d13973b92893c0a141163589ad290db151dc19f0 Validate config by checking the run commands (#2147)

1a1853f1b88d458594ef03671af15356b63ef9fd add llms.txt (#2120)

637614cfb33d498523cb539c1d6973c24e47c1a1 simplify llms.txt generation (#2123)

9f74e7d997d6a4978e8226471aef3f34f13c9bfb update token config in cog init actions workflow template (#2192)

0.14.0a16 Jan 2025pre-releaseBreaking

This is now unsupported and will receive no further updates. There are some breaking changes in the API that will be introduced with the release of th…

Support for concurrent predictions

This release introduces support for concurrent processing of predictions through the use of an async predict function.

To enable the feature add the new concurrency.max entry to your cog.yaml file:

concurrency:
  max: 32

And update your predictor to use the async def predict syntax:

class Predictor(BasePredictor):
    async def setup(self) -> None:
        print("async setup is also supported...")

    async def predict(self) -> str:
        print("async predict");
        return "hello world";

Cog will now process up to 32 predictions simultaneously, once at capacity subsequent predictions will return a 409 HTTP response.

Iterators

If your model is currently using Iterator or ConcatenateIterator it will need to be updated to use AsyncIterator or AsyncConcatenateIterator respectively.

from cog import AsyncConcatenateIterator, BasePredictor

class Predict(BasePredictor):
    async def predict(self) -> AsyncConcatenateIterator[str]:
        for fruit in ["apple", "banana", "orange"]:
            yield fruit

Migrating from 0.10.0a

An earlier fork of cog with concurrency support was published under the 0.10.0a release channel. This is now unsupported and will receive no further updates. There are some breaking changes in the API that will be introduced with the release of the 0.14.0 beta. This alpha release is backwards compatible and you will see deprecation warnings when calling the deprecated functions.

emit_metric(name, value) - this has been replaced by current_scope().record_metric(name, value)

[!NOTE] Note that the use of current_scope is still experimental and will output warnings to the console. To suppress these you can ignore the ExperimentalFeatureWarning:
import warnings
from cog import ExperimentalFeatureWarning
warnings.filterwarnings("ignore", category=ExperimentalFeatureWarning)

Known limitations

An async setup method cannot be used without an async predict method. Supported combinations are: sync setup/sync predict, async setup/async predict and sync setup/async predict.
File uploads will block the event loop. If your model outputs File or Path types these will currently block the event loop. This may be an issue for large file outputs and will be fixed in a future release.

Other Changes

Change torch vision to 0.20.0 for torch 2.5.0 cpu by @8W9aG in https://github.com/replicate/cog/pull/2074
Ignore files within a .git directory by @8W9aG in https://github.com/replicate/cog/pull/2087
Add fast build flag to cog by @8W9aG in https://github.com/replicate/cog/pull/2086
Make dockerfile generators abstract by @8W9aG in https://github.com/replicate/cog/pull/2088
Do not run a separate python install stage by @8W9aG in https://github.com/replicate/cog/pull/2094

Full Changelog: https://github.com/replicate/cog/compare/v0.13.6...v0.14.0-alpha1

cog

Release timeline

Releases

Changelog

Support for concurrent predictions

Iterators

Migrating from 0.10.0a

Known limitations

Changelog

Support for concurrent predictions

Iterators

Migrating from 0.10.0a

Known limitations

Other Changes

Changelog

Changelog

Changelog

Changelog

Changelog

Changelog

Changelog

Changelog

Changelog

Changelog

What's Changed

Changelog

Changelog

Changelog

Changelog

Changelog

Changelog

Changelog

What's Changed

What's Changed

What's Changed

Changelog

What's Changed

What's Changed

What's Changed

What's Changed

What's Changed

What's Changed

Changelog

Changelog

Changelog

Changelog

What's Changed

New Contributors

Changelog

Changelog

Changelog

Changelog

Changelog

What's Changed

New Contributors

What's Changed

New Contributors

What's Changed

New Contributors

What's Changed

New Contributors

Changelog

Changelog

What's Changed

New Contributors

What's Changed

What's Changed

Changelog