sgl-project/sglang · error · ValueError

material URI has data after base64 padding

Error message

material URI has data after base64 padding

What it means

Raised while streaming a base64/data material URI for MiniMax H3: after one or more '=' padding characters were seen, a non-'=' base64 character followed. The strict streaming decoder only allows up to two trailing '=' chars at the very end of the payload.

Source

Thrown at python/sglang/multimodal_gen/runtime/pipelines_core/stages/model_specific_stages/minimax_h3/material_io.py:619

        condition_index=condition_index,
        media_type=media_type,
    )
    total = 0
    encoded_size = 0
    padding = 0
    saw_padding = False
    encoded_chunk = bytearray()
    try:
        with partial_path.open("wb") as output:
            for value in _iter_base64_payload_bytes(uri, payload_start):
                encoded_size += 1
                if value == ord("="):
                    saw_padding = True
                    padding += 1
                    if padding > 2:
                        raise ValueError("material URI has invalid base64 padding")
                elif saw_padding:
                    raise ValueError("material URI has data after base64 padding")
                encoded_chunk.append(value)
                if len(encoded_chunk) == MINIMAX_H3_BASE64_DECODE_CHUNK_CHARS:
                    decoded_chunk = _decode_base64_chunk(encoded_chunk)
                    total += len(decoded_chunk)
                    output.write(decoded_chunk)
                    encoded_chunk.clear()
            if encoded_size == 0:
                raise ValueError("material URI base64 payload is empty")
            if encoded_size % 4 == 1 or (padding and encoded_size % 4):
                raise ValueError("material URI has an invalid base64 payload length")
            if encoded_chunk:
                encoded_chunk.extend(b"=" * (-len(encoded_chunk) % 4))
                decoded_chunk = _decode_base64_chunk(encoded_chunk)
                total += len(decoded_chunk)
                output.write(decoded_chunk)
        decoded_size = (encoded_size * 3) // 4 - padding
        if decoded_size <= 0:
            raise ValueError("material URI decoded payload is empty")

View on GitHub (pinned to 0132848349)

Solutions

1. Inspect the payload after the final '=' and strip anything following it
2. Regenerate the data URI with base64.b64encode() instead of hand-editing
3. Validate with base64.b64decode(payload, validate=True) before submitting

Example fix

// before
uri = f"data:image/png;base64,{b64}=extra"
// after
uri = f"data:image/png;base64,{b64}"

import base64
payload = uri.split(',', 1)[1] if uri.startswith('data:') else uri.split(':', 1)[1]
base64.b64decode(payload, validate=True)  # raises on data after padding

try:
    path = minimax_h3_localize_material_uri(batch, uri, ...)
except ValueError as e:
    if 'base64' in str(e):
        reject_request(f'bad material URI: {e}')

When it happens

Trigger: Calling minimax_h3_localize_material_uri with a data:/base64: URI whose base64 payload contains non-padding characters after '=' (e.g. 'aGVsbG8=9' or duplicated '=' mid-string).

Common situations: Hand-assembled or LLM-generated data URIs, string concatenation that appends query params after the padding, or trailing whitespace/CR-LF stripped incorrectly producing mid-payload '='.

Related errors

• tar material offset_data and size must be non-negative
• material URI has an invalid base64 payload
• material URI has invalid base64 padding
• condition URI must be a non-empty string
• No base64 image data found

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/5f2b0e5481c14465. Report an issue: GitHub.