opendatalab/MinerU · critical · FileNotFoundError

{} is not found.

Error message

{} is not found.

What it means

Raised by AnalysisConfig when the OCR weights path (made absolute first) does not exist. It is the top-level guard before either reading the YAML config or inferring the architecture from the weights filename.

Source

Thrown at mineru/model/utils/tools/infer/pytorchocr_utility.py:173

def read_network_config_from_yaml(yaml_path, char_num=None):
    if not os.path.exists(yaml_path):
        raise FileNotFoundError('{} is not existed.'.format(yaml_path))
    import yaml
    with open(yaml_path, encoding='utf-8') as f:
        res = yaml.safe_load(f)
    if res.get('Architecture') is None:
        raise ValueError('{} has no Architecture'.format(yaml_path))
    if res['Architecture']['Head']['name'] == 'MultiHead' and char_num is not None:
        res['Architecture']['Head']['out_channels_list'] = {
            'CTCLabelDecode': char_num,
            'SARLabelDecode': char_num + 2,
            'NRTRLabelDecode': char_num + 3
        }
    return res['Architecture']

def AnalysisConfig(weights_path, yaml_path=None, char_num=None):
    if not os.path.exists(os.path.abspath(weights_path)):
        raise FileNotFoundError('{} is not found.'.format(weights_path))

    if yaml_path is not None:
        return read_network_config_from_yaml(yaml_path, char_num=char_num)


def resize_img(img, input_size=600):
    """
    resize img and limit the longest side of the image to input_size
    """
    img = np.array(img)
    im_shape = img.shape
    im_size_max = np.max(im_shape[0:2])
    im_scale = float(input_size) / float(im_size_max)
    img = cv2.resize(img, None, None, fx=im_scale, fy=im_scale)
    return img


def str_count(s):

View on GitHub (pinned to 4fe4bde114)

Solutions

1. Verify with ls; correct the path or make it absolute.
2. Download the models first using mineru's model download commands.
3. Check MINERU_MODEL_SOURCE/model-dir env vars that determine where weights are looked up.

Example fix

# before
AnalysisConfig(weights_path="models/ocr_rec.pth")  # run from wrong cwd

# after
import os
AnalysisConfig(weights_path=os.path.abspath("models/ocr_rec.pth"))

import os
weights_path = os.path.abspath(weights_path)
if not os.path.isfile(weights_path):
    raise FileNotFoundError(f"weights not found: {weights_path}")

try:
    config = AnalysisConfig(weights_path, yaml_path)
except FileNotFoundError as e:
    raise SystemExit(f"model files missing, run download step: {e}") from e

When it happens

Trigger: Running the pytorchocr inference utility with a --weights path (or derived model path) that is wrong: typo, not-yet-downloaded model, relative path resolved against an unexpected cwd.

Common situations: Fresh environments where models were never downloaded; container images missing the models volume; scripts run from another directory making relative paths fail.

Related errors

• {} is not existed.
• {} is not existed.
• architecture {file_name} is not in arch_config.yaml
• Language {lang} not supported. Allowed values: {allowed_valu
• Missing middle json file: {middle_json_path}

AI-assisted analysis of opendatalab/MinerU@4fe4bde114 (2026-08-14). Data as JSON: /api/errors/37582a64ebf103fc. Report an issue: GitHub.