[Bug] Qwen2.5-VL prepare_inputs_for_generation does not pass second_per_grid_ts to get_rope_index

[Xqle]

System Info

When using Qwen2.5-VL for generation, the method prepare_inputs_for_generation calls get_rope_index without passing second_per_grid_ts.

def prepare_inputs_for_generation(
        self,
        input_ids,
        past_key_values=None,
        attention_mask=None,
        inputs_embeds=None,
        cache_position=None,
        position_ids=None,
        use_cache=True,
        pixel_values=None,
        pixel_values_videos=None,
        image_grid_thw=None,
        video_grid_thw=None,
        second_per_grid_ts=None,
        **kwargs,
    ):
        # Overwritten -- in specific circumstances we don't want to forward image inputs to the model

        model_inputs = super().prepare_inputs_for_generation(
            input_ids,
            past_key_values=past_key_values,
            attention_mask=attention_mask,
            inputs_embeds=inputs_embeds,
            cache_position=cache_position,
            position_ids=position_ids,
            pixel_values=pixel_values,
            pixel_values_videos=pixel_values_videos,
            image_grid_thw=image_grid_thw,
            video_grid_thw=video_grid_thw,
            second_per_grid_ts=second_per_grid_ts,
            use_cache=use_cache,
            **kwargs,
        )

        # Qwen2-5-VL position_ids are prepared with rope_deltas
        if position_ids is None:
            # Calculate RoPE index once per generation in the pre-fill stage only.
            # When compiling, we can't check tensor values thus we check only input length
            # It is safe to assume that `length!=1` means we're in pre-fill because compiled
            # models currently cannot do assisted decoding
            if cache_position[0] == 0 or self.model.rope_deltas is None:
                vision_positions, rope_deltas = self.model.get_rope_index(
                    model_inputs.get("input_ids", None),
                    image_grid_thw=image_grid_thw,
                    video_grid_thw=video_grid_thw,
                    attention_mask=attention_mask,
                    # ⚠ second_per_grid_ts is missing here
                )
                self.model.rope_deltas = rope_deltas
            # then use the prev pre-calculated rope-deltas to get the correct position ids
            elif "position_ids" in model_inputs:
                ...

        return model_inputs

This is inconsistent with the training setup, where second_per_grid_ts is provided in Qwen2_5_VLModel.forward:

    @auto_docstring
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Cache] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        pixel_values: Optional[torch.Tensor] = None,
        pixel_values_videos: Optional[torch.FloatTensor] = None,
        image_grid_thw: Optional[torch.LongTensor] = None,
        video_grid_thw: Optional[torch.LongTensor] = None,
        rope_deltas: Optional[torch.LongTensor] = None,
        cache_position: Optional[torch.LongTensor] = None,
        second_per_grid_ts: Optional[torch.Tensor] = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> Union[tuple, Qwen2_5_VLModelOutputWithPast]:
        r"""
        image_grid_thw (`torch.LongTensor` of shape `(num_images, 3)`, *optional*):
            The temporal, height and width of feature shape of each image in LLM.
        video_grid_thw (`torch.LongTensor` of shape `(num_videos, 3)`, *optional*):
            The temporal, height and width of feature shape of each video in LLM.
        rope_deltas (`torch.LongTensor` of shape `(batch_size, )`, *optional*):
            The rope index difference between sequence length and multimodal rope.
        second_per_grid_ts (`torch.Tensor` of shape `(num_videos)`, *optional*):
            The time interval (in seconds) for each grid along the temporal dimension in the 3D position IDs.
        """

        ...

        if position_ids is None:
            # Calculate RoPE index once per generation in the pre-fill stage only.
            # When compiling, we can't check tensor values thus we check only input length
            # It is safe to assume that `length!=1` means we're in pre-fill because compiled
            # models currently cannot do asssisted decoding
            prefill_compiled_stage = is_torchdynamo_compiling() and (
                (input_ids is not None and input_ids.shape[1] != 1)
                or (inputs_embeds is not None and inputs_embeds.shape[1] != 1)
            )
            prefill_noncompiled_stage = not is_torchdynamo_compiling() and (
                (cache_position is not None and cache_position[0] == 0)
                or (past_key_values is None or past_key_values.get_seq_length() == 0)
            )
            if (prefill_compiled_stage or prefill_noncompiled_stage) or self.rope_deltas is None:
                position_ids, rope_deltas = self.get_rope_index(
                    input_ids,
                    image_grid_thw,
                    video_grid_thw,
                    second_per_grid_ts=second_per_grid_ts,  # Here passing the second_per_grid_ts
                    attention_mask=attention_mask,
                )
                self.rope_deltas = rope_deltas
            else:
                batch_size, seq_length, _ = inputs_embeds.shape
                position_ids = torch.arange(seq_length, device=inputs_embeds.device)
                position_ids = position_ids.view(1, 1, -1).expand(3, batch_size, -1)
                if cache_position is not None:
                    delta = (cache_position[0] + self.rope_deltas).to(inputs_embeds.device)
                else:
                    delta = torch.zeros((batch_size, seq_length), device=inputs_embeds.device)
                delta = delta.repeat_interleave(batch_size // delta.shape[0], dim=1)
                position_ids = position_ids + delta.to(position_ids.device)

        ...

        return output if return_dict else output.to_tuple()

As a result, the model may compute incorrect RoPE indices for video inputs, especially when variable frame rates or temporal scaling are involved.

Who can help?

No response

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

1. Load Qwen2.5-VL from Hugging Face:

from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-VL")
processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL")

2. Run generation with video inputs.
3. Check prepare_inputs_for_generation:

vision_positions, rope_deltas = self.model.get_rope_index(
    model_inputs.get("input_ids", None),
    image_grid_thw=image_grid_thw,
    video_grid_thw=video_grid_thw,
    attention_mask=attention_mask,
    # ⚠ second_per_grid_ts is missing here
)

Expected behavior

• second_per_grid_ts should be passed into get_rope_index during generation, to align with training behavior.
• This ensures temporal RoPE indices are computed consistently between training and inference.