Pydantic >2 and optional image Inputs

[jschoormans]

Hi,
I'm developing a cog model that requires a dependeny, which needs pydantic >2.
Now, I'm noticing an issue with optional image inputs. The model returns an error in that case. I'll provide a minimal working example:

from cog import BasePredictor, Path, Input

class Predictor(BasePredictor):
    def predict(
        self,
        test_image: Path = Input(description="Test image", default=None),
    ) -> Path:
        """Run a single prediction on the model"""

        return Path("./hello.webp")

build:
  python_version: "3.11"

  python_packages:
    - "pydantic==2.1.0"


predict: "predict.py:Predictor"

Runs perfectly fine with cog predict -i test_image=@...; however with no input it gives this error:

{"logger": "uvicorn.error", "timestamp": "2025-03-17T11:30:06.489188Z", "exception": "Traceback (most recent call last):\n  File \"/usr/local/lib/python3.11/site-packages/uvicorn/protocols/http/httptools_impl.py\", line 401, in run_asgi\n    result = await app(  # type: ignore[func-returns-value]\n             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/uvicorn/middleware/proxy_headers.py\", line 70, in __call__\n    return await self.app(scope, receive, send)\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/fastapi/applications.py\", line 1054, in __call__\n    await super().__call__(scope, receive, send)\n  File \"/usr/local/lib/python3.11/site-packages/starlette/applications.py\", line 112, in __call__\n    await self.middleware_stack(scope, receive, send)\n  File \"/usr/local/lib/python3.11/site-packages/starlette/middleware/errors.py\", line 187, in __call__\n    raise exc\n  File \"/usr/local/lib/python3.11/site-packages/starlette/middleware/errors.py\", line 165, in __call__\n    await self.app(scope, receive, _send)\n  File \"/usr/local/lib/python3.11/site-packages/starlette/middleware/exceptions.py\", line 62, in __call__\n    await wrap_app_handling_exceptions(self.app, conn)(scope, receive, send)\n  File \"/usr/local/lib/python3.11/site-packages/starlette/_exception_handler.py\", line 53, in wrapped_app\n    raise exc\n  File \"/usr/local/lib/python3.11/site-packages/starlette/_exception_handler.py\", line 42, in wrapped_app\n    await app(scope, receive, sender)\n  File \"/usr/local/lib/python3.11/site-packages/starlette/routing.py\", line 714, in __call__\n    await self.middleware_stack(scope, receive, send)\n  File \"/usr/local/lib/python3.11/site-packages/starlette/routing.py\", line 734, in app\n    await route.handle(scope, receive, send)\n  File \"/usr/local/lib/python3.11/site-packages/starlette/routing.py\", line 288, in handle\n    await self.app(scope, receive, send)\n  File \"/usr/local/lib/python3.11/site-packages/starlette/routing.py\", line 76, in app\n    await wrap_app_handling_exceptions(app, request)(scope, receive, send)\n  File \"/usr/local/lib/python3.11/site-packages/starlette/_exception_handler.py\", line 53, in wrapped_app\n    raise exc\n  File \"/usr/local/lib/python3.11/site-packages/starlette/_exception_handler.py\", line 42, in wrapped_app\n    await app(scope, receive, sender)\n  File \"/usr/local/lib/python3.11/site-packages/starlette/routing.py\", line 73, in app\n    response = await f(request)\n               ^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/fastapi/routing.py\", line 301, in app\n    raw_response = await run_endpoint_function(\n                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/fastapi/routing.py\", line 212, in run_endpoint_function\n    return await dependant.call(**values)\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/cog/server/http.py\", line 364, in predict\n    return await _predict(\n           ^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/cog/server/http.py\", line 470, in _predict\n    _ = response_type(**response_object)\n        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/pydantic/main.py\", line 159, in __init__\n    __pydantic_self__.__pydantic_validator__.validate_python(data, self_instance=__pydantic_self__)\n  File \"/usr/local/lib/python3.11/site-packages/cog/types.py\", line 205, in validate\n    filename=get_filename(value),\n             ^^^^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/cog/types.py\", line 405, in get_filename\n    basename = urllib.parse.unquote_plus(basename)\n               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/urllib/parse.py\", line 808, in unquote_plus\n    string = string.replace('+', ' ')\n             ^^^^^^^^^^^^^^^^^^^^^^^^\nTypeError: a bytes-like object is required, not 'str'", "severity": "ERROR", "message": "Exception in ASGI application\n"}

I'm on cog version 0.14.1

Thanks!!

[whistler]

I have the same issue, inspecting the code, the problem seems to be that if you are not using a cog base image, cog is installed after pip packages: https://github.com/replicate/cog/blob/main/pkg/dockerfile/standard_generator.go#L181. During cog installation, it installs pydantic<2.

[aron]

Hi folks, could you please test this with cog 0.14.3, we've made some changes to better support optional inputs. The syntax would now be:

class Predictor(BasePredictor):
    def predict(
        self,
        test_image: Path | None = Input(description="Test image", default=None),
    ) -> Path:
        ...

https://github.com/replicate/cog/releases/tag/v0.14.3

[jschoormans]

Hi @aron,

Great to hear! I still have some issues with this.

I'm now on 0.14.3 on Ubuntu.
cog --version cog version 0.14.3 (built 2025-03-20T19:52:21Z)

Using this optional input I get the following error when running predict:

TypeError: Unsupported input type NoneType for parameter test_image. Supported input types are: str, int, float, bool, cog.File, cog.Path, cog.Secret, or a Union or List of those types.

This happens with Python 3.10 or 3.11 in cog (without defining another pydantic version in cog yaml)

`build:
python_version: "3.10"

python_packages:
- "pillow"

predict: "predict.py:Predictor"
`

`from cog import BasePredictor, Path, Input
from PIL import Image

class Predictor(BasePredictor):
def predict(
self,
test_image: Path | None = Input(description="Test image", default=None),
) -> Path:
"""Run a single prediction on the model"""
im = Image.new("RGB", (100, 100), color="red")
im.save(Path("./hello.webp"))
return Path("./hello.webp")`

[aron]

Ah, it looks like we only tested this with Optional. You can work around this for now by using that:

from typing import Optional
from cog import BasePredictor, Path, Input

class Predictor(BasePredictor):
  def predict(
    self,
    test_image: Optional[Path] = Input(description="Test image", default=None),
  ) -> Path:
    ...

I saw you've opened a PR, thank you! We'll get that looked at.

[jschoormans]

Ok, perfect. That works in all cases (pydantic 1 and 2, with and without an input)! Thanks!

[8W9aG]

I've got a PR that I believe solves this: #2224