PaddlePaddle/PaddleOCR · error · ValueError

Invalid input name {repr(name)} found in `dynamic_shape_inpu

Error message

Invalid input name {repr(name)} found in `dynamic_shape_input_data`

What it means

During TRT engine building, dynamic_shape_input_data optionally provides real calibration/shape data per input. Its keys are validated against predictor.get_input_names(); a key that matches no model input raises this ValueError. It fires after the dynamic_shapes checks, and guards against feeding sample data to a nonexistent tensor, which would silently be ignored or crash the engine build.

Source

Thrown at tools/infer/utility.py:482

        config.disable_mkldnn()
        config.disable_glog_info()
        return inference.create_predictor(config)

    dynamic_shape_input_data = dynamic_shape_input_data or {}

    predictor = _get_predictor(pp_model_file, pp_params_file)
    input_names = predictor.get_input_names()
    for name in dynamic_shapes:
        if name not in input_names:
            raise ValueError(
                f"Invalid input name {repr(name)} found in `dynamic_shapes`"
            )
    for name in input_names:
        if name not in dynamic_shapes:
            raise ValueError(f"Input name {repr(name)} not found in `dynamic_shapes`")
    for name in dynamic_shape_input_data:
        if name not in input_names:
            raise ValueError(
                f"Invalid input name {repr(name)} found in `dynamic_shape_input_data`"
            )

    trt_inputs = []
    for name, candidate_shapes in dynamic_shapes.items():
        # XXX: Currently we have no way to get the data type of the tensor
        # without creating an input handle.
        handle = predictor.get_input_handle(name)
        dtype = _pd_dtype_to_np_dtype(handle.type())
        min_shape, opt_shape, max_shape = candidate_shapes
        if name in dynamic_shape_input_data:
            min_arr = np.array(dynamic_shape_input_data[name][0], dtype=dtype).reshape(
                min_shape
            )
            opt_arr = np.array(dynamic_shape_input_data[name][1], dtype=dtype).reshape(
                opt_shape
            )
            max_arr = np.array(dynamic_shape_input_data[name][2], dtype=dtype).reshape(

View on GitHub (pinned to 2661c7c0ef)

Solutions

1. Make the keys of trt_dynamic_shapes_input_data (or the function argument) exactly match predictor.get_input_names()
2. Remove stale input-data entries entirely if calibration data is not required
3. Regenerate the config from the model's own inference.yml

Example fix

# before
trt_dynamic_shapes_input_data: {x: [[[...]]]}   # input actually 'images'

# after
trt_dynamic_shapes_input_data:
  images: [[[1,3,48,640]]]

input_names = set(predictor.get_input_names())
assert set(dynamic_shape_input_data) <= input_names, \
    f'dynamic_shape_input_data keys must be actual inputs, got extra: {set(dynamic_shape_input_data) - input_names}'

try:
    _convert_trt(dynamic_shapes, model_file, params_file, dynamic_shape_input_data=..., ...)
except ValueError as e:
    if 'dynamic_shape_input_data' in str(e):
        dynamic_shape_input_data = {k: v for k, v in dynamic_shape_input_data.items() if k in input_names}
        _convert_trt(..., dynamic_shape_input_data=dynamic_shape_input_data, ...)
    raise

When it happens

Trigger: Passing dynamic_shape_input_data={'x': [...]} when the model input is named differently ('images', 'input'); reusing the input-data section of an inference.yml with another model.

Common situations: Same root cause as the dynamic_shapes mismatches: yml copied between det/rec/e2e models whose tensor names differ, or renamed inputs across export versions.

Related errors

• Invalid input name {repr(name)} found in `dynamic_shapes`
• Input name {repr(name)} not found in `dynamic_shapes`
• Configuration Error: 'trt_dynamic_shapes' must be defined in
• Unsupported data type: {pd_dtype}

AI-assisted analysis of PaddlePaddle/PaddleOCR@2661c7c0ef (2026-08-14). Data as JSON: /api/errors/51d7ff19204e791e. Report an issue: GitHub.