sgl-project/sglang · critical · ValueError

hidden_size must be positive.

Error message

hidden_size must be positive.

What it means

MiniMax H3's constructor requires hidden_size > 0 since it sizes input patch projections, AdaLN layers, and TP-sharded linear weights from it. A non-positive hidden_size indicates a broken arch config and the model refuses to build.

Source

Thrown at python/sglang/multimodal_gen/runtime/models/dits/minimax_h3.py:1809

        return (
            self.adaln_cache is None
            and get_tp_world_size() > 1
            and not torch.compiler.is_compiling()
            and not envs.SGLANG_CACHE_DIT_ENABLED
            and not hasattr(self, "_sglang_cache_dit_adapter")
            and not is_layerwise_offloaded_module(self)
            and all(type(block) is MiniMaxH3DiTBlock for block in self.blocks)
        )

    def _validate_tp_config(
        self, *, arch: MiniMaxH3DiTArchConfig, tp_size: int
    ) -> None:
        if tp_size <= 0:
            raise ValueError("TP size must be positive.")
        if arch.num_attention_heads <= 0:
            raise ValueError("num_attention_heads must be positive.")
        if arch.hidden_size <= 0:
            raise ValueError("hidden_size must be positive.")
        if arch.attention_head_dim <= 0:
            raise ValueError("attention_head_dim must be positive.")
        if arch.ffn_hidden_size <= 0:
            raise ValueError("ffn_hidden_size must be positive.")
        for name, value in (
            ("num_attention_heads", arch.num_attention_heads),
            ("hidden_size", arch.hidden_size),
            ("ffn_hidden_size", arch.ffn_hidden_size),
            ("time_embed_hidden_size", arch.time_embed_hidden_size),
            ("adaln_out_features", arch.adaln_out_features),
            ("final_adaln_out_features", arch.final_adaln_out_features),
            ("video_patch_output_dim", arch.latents_dim * math.prod(arch.patch_size)),
            ("audio_patch_output_dim", arch.audio_latents_dim),
        ):
            if value % tp_size:
                raise ValueError(
                    f"MiniMax H3 {name}={value} must be divisible by "
                    f"TP size {tp_size}."

View on GitHub (pinned to 0132848349)

Solutions

1. Print/validate arch.hidden_size before constructing the model
2. Set it to the checkpoint's true embedding width (e.g. 3072)
3. Re-run the checkpoint converter and confirm all size fields are populated

Example fix

# before
arch.hidden_size = 0
# after
arch.hidden_size = cfg_json["hidden_size"]  # e.g. 3072

assert arch.hidden_size > 0

def hidden_ok(arch) -> bool:
    return getattr(arch, "hidden_size", 0) > 0

When it happens

Trigger: Passing a MiniMaxH3DiTArchConfig (or deserialized JSON config) where hidden_size is 0, negative, or omitted and defaulted to 0.

Common situations: Custom model configs authored from scratch; partial checkpoint conversion that forgets hidden_size; refactors that rename to dim/model_dim and leave hidden_size unset.

Related errors

• num_attention_heads must be positive.
• attention_head_dim must be positive.
• ffn_hidden_size must be positive.
• TP size must be positive.
• Invalid threshold_type for topk: {threshold_type}. Choose 'q

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/fa2714a838d31149. Report an issue: GitHub.