sgl-project/sglang · error · ValueError

input_ids must be 1-D, got {list(input_ids.shape)}

Error message

input_ids must be 1-D, got {list(input_ids.shape)}

What it means

encode_ids requires a 1-D input_ids tensor (a single flattened sequence). Passing batched (2-D) or higher-rank token id tensors raises this immediately.

Source

Thrown at python/sglang/multimodal_gen/runtime/models/encoders/minimax_h3_qwen3vl.py:339

            output_hidden_states=False,
            return_dict=True,
            use_cache=False,
            **kwargs,
        )
        return BaseEncoderOutput(last_hidden_state=outputs.last_hidden_state)

    @torch.no_grad()
    def encode_ids(
        self,
        input_ids: torch.Tensor,
        *,
        pixel_values: torch.Tensor | None = None,
        image_grid_thw: torch.Tensor | None = None,
        pixel_values_videos: torch.Tensor | None = None,
        video_grid_thw: torch.Tensor | None = None,
    ) -> torch.Tensor:
        if input_ids.dim() != 1:
            raise ValueError(f"input_ids must be 1-D, got {list(input_ids.shape)}")
        if (pixel_values is None) != (image_grid_thw is None):
            raise ValueError("pixel_values and image_grid_thw must be given together")
        if (pixel_values_videos is None) != (video_grid_thw is None):
            raise ValueError(
                "pixel_values_videos and video_grid_thw must be given together"
            )

        host_ids = input_ids.to(device="cpu", dtype=torch.long)[None]
        host_image_grid_thw = (
            image_grid_thw.to(device="cpu", dtype=torch.long)
            if image_grid_thw is not None
            else None
        )
        host_video_grid_thw = (
            video_grid_thw.to(device="cpu", dtype=torch.long)
            if video_grid_thw is not None
            else None
        )

View on GitHub (pinned to 0132848349)

Solutions

1. Squeeze/flatten: input_ids = input_ids.reshape(-1) or input_ids[0] before calling encode_ids
2. Loop over the batch, calling encode_ids once per sequence

Example fix

# before
out = encoder.encode_ids(tokenizer(...)["input_ids"])  # shape [1, N]
# after
ids = tokenizer(...)["input_ids"].reshape(-1)
out = encoder.encode_ids(ids)

assert input_ids.dim() == 1, f"need 1-D ids, got {tuple(input_ids.shape)}"

def is_flat_ids(t: torch.Tensor) -> bool:
    return isinstance(t, torch.Tensor) and t.dim() == 1

When it happens

Trigger: Calling encode_ids(input_ids) with input_ids.dim() != 1, e.g. a [1, seq_len] batched tensor straight from a tokenizer with return_tensors='pt'.

Common situations: Feeding tokenizer output shaped (batch, seq) without squeezing; adapting code from batched encoders like Qwen3VLForConditionalGeneration which use 2-D inputs.

Related errors

• x must be [1, S, C], got {list(x.shape)}
• pixel_values and image_grid_thw must be given together
• pixel_values_videos and video_grid_thw must be given togethe
• unexpected hidden shape {list(hidden.shape)}, expected {expe
• fl2va requires first_frame, last_frame, or both

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/65b799c2af9b2400. Report an issue: GitHub.