PaddlePaddle/PaddleOCR · error · ValueError

Invalid precision: {precision}. Supported values are: {SUPPO

Error message

Invalid precision: {precision}. Supported values are: {SUPPORTED_PRECISION_LIST}.

What it means

ValueError raised by parse_common_args when precision is not in SUPPORTED_PRECISION_LIST (['fp32', 'fp16']). Unlike engine, precision is always checked even at its default, so an invalid value anywhere in the config chain fails immediately.

Source

Thrown at paddleocr/_common_args.py:67

        "enable_cinn": DEFAULT_USE_CINN,
    }

    unknown_names = kwargs.keys() - default_vals.keys()
    for name in unknown_names:
        raise ValueError(f"Unknown argument: {name}")

    kwargs = {**default_vals, **kwargs}

    if (
        kwargs["engine"] is not None
        and kwargs["engine"] not in SUPPORTED_INFERENCE_ENGINE_LIST
    ):
        raise ValueError(
            f"Invalid engine: {kwargs['engine']}. Supported values are: {SUPPORTED_INFERENCE_ENGINE_LIST}."
        )

    if kwargs["precision"] not in SUPPORTED_PRECISION_LIST:
        raise ValueError(
            f"Invalid precision: {kwargs['precision']}. Supported values are: {SUPPORTED_PRECISION_LIST}."
        )

    kwargs["use_pptrt"] = kwargs.pop("use_tensorrt")
    kwargs["pptrt_precision"] = kwargs.pop("precision")

    return kwargs


def _build_paddle_static_engine_config(common_args, device_type):
    cfg = {}
    if device_type == "gpu":
        if common_args["use_pptrt"]:
            if common_args["pptrt_precision"] == "fp32":
                cfg["run_mode"] = "trt_fp32"
            else:
                assert common_args["pptrt_precision"] == "fp16", common_args[
                    "pptrt_precision"

View on GitHub (pinned to 2661c7c0ef)

Solutions

1. Use exactly 'fp32' or 'fp16' (lowercase)
2. For other numeric formats on TensorRT paths, check whether your version supports them via SUPPORTED_PRECISION_LIST before passing

Example fix

# before
ocr = PaddleOCR(precision='float16', use_tensorrt=True)
# after
ocr = PaddleOCR(precision='fp16', use_tensorrt=True)

from paddleocr._constants import SUPPORTED_PRECISION_LIST

if precision not in SUPPORTED_PRECISION_LIST:
    raise ValueError(f'precision must be one of {SUPPORTED_PRECISION_LIST}')

from typing import Literal

Precision = Literal['fp32', 'fp16']

def is_precision(v: str) -> 'TypeGuard[Precision]':
    return v in ('fp32', 'fp16')

When it happens

Trigger: PaddleOCR(precision='int8'), precision='float16', or precision='FP16' (case-sensitive check).

Common situations: Coming from TensorRT configs where 'int8'/'bf16' are valid; assuming case-insensitivity; abbreviating float16 as fp16 correctly but trying bfloat16.

Related errors

• worker mode does not support a custom fetch implementation.
• Conflicting values provided for ${label}: ${aliases.join(",
• ${modelRole} model selection must define model_name.
• OCR model selection must define both detection and recogniti
• worker must be a boolean or an options object.

AI-assisted analysis of PaddlePaddle/PaddleOCR@2661c7c0ef (2026-08-14). Data as JSON: /api/errors/972d026b0ec5ae78. Report an issue: GitHub.