PaddlePaddle/PaddleOCR · error · ValueError

mode can only be one of ['lite', 'large'], but received {}

Error message

mode can only be one of ['lite', 'large'], but received {}

What it means

RSEFPN (the DB-detection FPN in db_fpn.py) selects its pan-head convolution type from mode: 'lite' uses depthwise-separable DSConv, 'large' uses full nn.Conv2D. Any other string fails the elif chain and raises this ValueError at construction time.

Source

Thrown at ppocr/modeling/necks/db_fpn.py:449

class LKPAN(nn.Layer):
    def __init__(self, in_channels, out_channels, mode="large", **kwargs):
        super(LKPAN, self).__init__()
        self.out_channels = out_channels
        weight_attr = paddle.nn.initializer.KaimingUniform()

        self.ins_conv = nn.LayerList()
        self.inp_conv = nn.LayerList()
        # pan head
        self.pan_head_conv = nn.LayerList()
        self.pan_lat_conv = nn.LayerList()

        if mode.lower() == "lite":
            p_layer = DSConv
        elif mode.lower() == "large":
            p_layer = nn.Conv2D
        else:
            raise ValueError(
                "mode can only be one of ['lite', 'large'], but received {}".format(
                    mode
                )
            )

        for i in range(len(in_channels)):
            self.ins_conv.append(
                nn.Conv2D(
                    in_channels=in_channels[i],
                    out_channels=self.out_channels,
                    kernel_size=1,
                    weight_attr=ParamAttr(initializer=weight_attr),
                    bias_attr=False,
                )
            )

            self.inp_conv.append(
                p_layer(

View on GitHub (pinned to 2661c7c0ef)

Solutions

1. Set Neck.Mode to "lite" (DSConv, faster/smaller) or "large" (Conv2D, heavier) in the detection config
2. Check for stray whitespace/case in the Mode string (' lite'.lower() != 'lite' only if leading space — trim it)

Example fix

# before (config yaml)
Neck:
  name: RSEFPN
  Mode: 'small'
# after
Neck:
  name: RSEFPN
  Mode: 'lite'

mode = str(neck_cfg['Mode']).strip().lower()
assert mode in ('lite', 'large'), f"Mode must be 'lite' or 'large', got {neck_cfg['Mode']!r}"

def valid_fpn_mode(mode) -> bool:
    return isinstance(mode, str) and mode.strip().lower() in ('lite', 'large')

When it happens

Trigger: Building a DB/RSE detection model whose Neck config has Mode set to something other than 'lite' or 'large' (case-insensitive), e.g. 'small', 'base', or a typo like 'Lite ' (note: 'LITE'.lower()=='lite' so only exact-after-lower strings pass).

Common situations: Copying a detection yaml and changing Mode to match a backbone scale naming ('tiny'/'base'); older configs that used different mode names; typos in hand-written Neck sections.

Related errors

• embed_dim must be divisible by num_heads (got `embed_dim`: {
• DilatedReparamBlock requires kernel_size in [5,7,9,11,13], b
• Please set inference model dir in Global.inference_model or
• "out_size" must be an integer or tuple of integers
• Sliding window is currently only implemented for causal mask

AI-assisted analysis of PaddlePaddle/PaddleOCR@2661c7c0ef (2026-08-14). Data as JSON: /api/errors/a3610c54536b0e0d. Report an issue: GitHub.