Modular converter ignores my `Config` and my `ModelOutput` classes

[tonywu71]

System Info

• transformers version: 4.46.0.dev0
• Platform: macOS-15.0-arm64-arm-64bit
• Python version: 3.11.6
• Huggingface_hub version: 0.25.1
• Safetensors version: 0.4.5
• Accelerate version: 0.34.2
• Accelerate config: not found
• PyTorch version (GPU?): 2.4.1 (False)
• Tensorflow version (GPU?): not installed (NA)
• Flax version (CPU?/GPU?/TPU?): not installed (NA)
• Jax version: not installed
• JaxLib version: not installed
• Using distributed or parallel set-up in script?: no

Who can help?

• @ArthurZucker
• @yonigozlan (reviewer for my ColPali integration PR Add ColPali to 🤗 transformers #33736)

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

Given the src/transformers/models/colpali/modular_colpali.py file below:

# coding=utf-8
# Copyright 2024 The HuggingFace Inc. team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.


from dataclasses import dataclass
from typing import ClassVar, List, Optional, Tuple, Union

import torch
import torch.utils.checkpoint
from torch import nn

from ...cache_utils import Cache
from ...utils import (
    ModelOutput,
    add_start_docstrings,
    add_start_docstrings_to_model_forward,
    is_flash_attn_2_available,
    logging,
    replace_return_docstrings,
)
from ..paligemma import (
    PaliGemmaConfig,
    PaliGemmaForConditionalGeneration,
)


if is_flash_attn_2_available():
    from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input  # noqa


logger = logging.get_logger(__name__)


class ColPaliConfig(PaliGemmaConfig):
    r"""
    This is the configuration class to store the configuration of a [`ColPaliForRetrieval`]. It is used to instantiate an
    ColPaliForRetrieval according to the specified arguments, defining the model architecture.

    The ColPali config is very similar to [`PaligemmaConfig`], but with an extra attribute defining the embedding dimension.

    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
    documentation from [`PretrainedConfig`] for more information.


    Args:
        vision_config (`PaliGemmaVisionConfig`,  *optional*):
            Custom vision config or dict
        text_config (`Union[AutoConfig, dict]`, *optional*):
            The config object of the text backbone. Can be any of `LlamaConfig` or `MistralConfig`.
        ignore_index (`int`, *optional*, defaults to -100):
            The ignore index for the loss function.
        image_token_index (`int`, *optional*, defaults to 256000):
            The image token index to encode the image prompt.
        vocab_size (`int`, *optional*, defaults to 257152):
            Vocabulary size of the PaliGemmamodel. Defines the number of different tokens that can be represented by the
            `inputs_ids` passed when calling [`~PaliGemmaForConditionalGeneration`]
        projection_dim (`int`, *optional*, defaults to 2048):
            Dimension of the multimodal projection space.
        hidden_size (`int`, *optional*, defaults to 2048):
            Dimension of the hidden layer of the Language model.
        embedding_dim (`int`, *optional*, defaults to 128):
            Dimension of the multi-vector embeddings produced by the model.

    Example:

    ```python
    from transformers.models.colpali import ColPaliConfig, ColPaliForRetrieval

    config = ColPaliConfig()
    model = ColPaliForRetrieval(config)
    ```
    """

    def __init__(
        self,
        vision_config=None,
        text_config=None,
        ignore_index=-100,
        image_token_index=256000,
        vocab_size=257152,
        projection_dim=2048,
        hidden_size=2048,
        embedding_dim: int = 128,
        **kwargs,
    ):
        super().__init__(
            vision_config=vision_config,
            text_config=text_config,
            ignore_index=ignore_index,
            image_token_index=image_token_index,
            vocab_size=vocab_size,
            projection_dim=projection_dim,
            hidden_size=hidden_size,
            **kwargs,
        )
        self.model_type = "colpali"
        self.is_composition = False
        self.embedding_dim = embedding_dim


@dataclass
class ColPaliForRetrievalOutput(ModelOutput):
    """
    Base class for ColPali embeddings output.

    Args:
        embeddings (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
            The embeddings of the model.
        loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided):
            Language modeling loss (for next-token prediction).
        past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
            Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
            `(batch_size, num_heads, sequence_length, embed_size_per_head)`)

            Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see
            `past_key_values` input) to speed up sequential decoding.
        hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
            Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.

            Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.
        attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
            sequence_length)`.

            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
            heads.
        image_hidden_states (`torch.FloatTensor`, *optional*):
            A `torch.FloatTensor` of size `(batch_size, num_images, sequence_length, hidden_size)`.
            image_hidden_states of the model produced by the vision encoder after projecting last hidden state.
    """

    embeddings: torch.Tensor = None
    loss: Optional[torch.FloatTensor] = None
    past_key_values: Optional[Union[List[torch.FloatTensor], Cache]] = None
    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
    attentions: Optional[Tuple[torch.FloatTensor]] = None
    image_hidden_states: Optional[torch.FloatTensor] = None


@add_start_docstrings(
    """
    ColPali leverages Vision Language Models (VLMs) to construct efficient multi-vector embeddings in the visual space for document retrieval.
    By feeding the ViT output patches from PaliGemma-3B to a linear projection, we create a multi-vector representation of documents. The model
    is trained to maximize the similarity between these document embeddings and the query embeddings, following the ColBERT method.

    Using ColPali removes the need for potentially complex and brittle layout recognition and OCR pipelines with a single model that can take into account
    both the textual and visual content (layout, charts, ...) of a document.

    ColPali was introduced in the following paper: [*ColPali: Efficient Document Retrieval with Vision Language Models*](https://arxiv.org/abs/2407.01449).

    Resources:
    - A blog post detailing ColPali, a vision retrieval model, can be found [here](https://huggingface.co/blog/manu/colpali). 📝
    - The code for training ColPali and for the `colpali-engine` package can be found [here](https://github.com/illuin-tech/colpali). 🌎
    - Cookbooks to fine-tune ColPali (with optional quantization), generate similarity maps, ... can be found [here](https://github.com/tonywu71/colpali-cookbooks). 📚

    Adapted from [`colpali-engine==0.3.0`](https://github.com/illuin-tech/colpali/releases/tag/v0.3.0).
    """
)
class ColPaliForRetrieval(PaliGemmaForConditionalGeneration):
    main_input_name: ClassVar[str] = "doc_input_ids"  # transformers-related

    def __init__(self, config: ColPaliConfig):
        super().__init__(config=config)

        self.embedding_dim = self.config.embedding_dim
        self.custom_text_proj = nn.Linear(self.config.text_config.hidden_size, self.embedding_dim)

        if self.language_model._tied_weights_keys is not None:
            self._tied_weights_keys = [f"model.language_model.{k}" for k in self.language_model._tied_weights_keys]

        self.post_init()

    @add_start_docstrings_to_model_forward(
        """
        Args:
        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
            it.
            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.
            [What are input IDs?](../glossary#input-ids)
        pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, image_size, image_size)):
            The tensors corresponding to the input images. Pixel values can be obtained using
            [`AutoImageProcessor`]. See [`SiglipImageProcessor.__call__`] for details ([]`PaliGemmaProcessor`] uses
            [`SiglipImageProcessor`] for processing images). If none, ColPali will only process text (query embeddings).
        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
            - 1 for tokens that are **not masked**,
            - 0 for tokens that are **masked**.
            [What are attention masks?](../glossary#attention-mask)
            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.
            If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see
            `past_key_values`).
            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
            information on the default strategy.
            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.
        """
    )
    @replace_return_docstrings(output_type=ColPaliForRetrievalOutput, config_class="ColPaliConfig")
    def forward(
        self,
        input_ids: torch.LongTensor = None,
        pixel_values: torch.FloatTensor = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Union[List[torch.FloatTensor], Cache]] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        cache_position: Optional[torch.LongTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        num_logits_to_keep: int = 0,
    ) -> Union[Tuple, ColPaliForRetrievalOutput]:
        r"""
        Returns:
        """
        vlm_outputs = super().forward(
            input_ids=input_ids,
            pixel_values=pixel_values,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            token_type_ids=token_type_ids,
            cache_position=cache_position,
            inputs_embeds=inputs_embeds,
            labels=labels,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=True,
            return_dict=True,
            num_logits_to_keep=num_logits_to_keep,
        )
        last_hidden_states = vlm_outputs.hidden_states[-1]  # (batch_size, sequence_length, hidden_size)
        proj = self.custom_text_proj(last_hidden_states)  # (batch_size, sequence_length, dim)

        # L2 normalization
        embeddings = proj / proj.norm(dim=-1, keepdim=True)  # (batch_size, sequence_length, dim)

        embeddings = embeddings * attention_mask.unsqueeze(-1)  # (batch_size, sequence_length, dim)

        if not return_dict:
            return (embeddings,) + vlm_outputs

        return ColPaliForRetrievalOutput(
            embeddings=embeddings,
            past_key_values=vlm_outputs.past_key_values,
            hidden_states=vlm_outputs.hidden_states,
            attentions=vlm_outputs.attentions,
            image_hidden_states=vlm_outputs.image_hidden_states,
        )

    def resize_token_embeddings(
        self,
        new_num_tokens: Optional[int] = None,
        pad_to_multiple_of=None,
    ) -> nn.Embedding:
        model_embeds = self.language_model.resize_token_embeddings(new_num_tokens, pad_to_multiple_of)

        # Update vocab size
        self.config.text_config.vocab_size = model_embeds.num_embeddings
        self.config.vocab_size = model_embeds.num_embeddings
        self.vocab_size = model_embeds.num_embeddings

        return model_embeds

After running:

python utils/modular_model_converter.py --files_to_parse src/transformers/models/colpali/modular_colpali.py

I obtain the following src/transformers/models/colpali/modeling_colpali.py file.

#           🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
#               This file was automatically generated from <path_to_modular_file.py>.
#         Do NOT edit this file manually as any edits will be overwritten by the generation of
#         the file from the modular. If any change should be done, please apply the change to the
#                           modular_xxx.py file directly. One of our CI enforces this
#           🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# coding=utf-8
# Copyright 2024 The HuggingFace Inc. team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.


from dataclasses import dataclass
from typing import ClassVar, List, Optional, Tuple, Union

import torch
import torch.utils.checkpoint
from torch import nn

from ...cache_utils import Cache
from ...utils import (
    ModelOutput,
    add_start_docstrings,
    add_start_docstrings_to_model_forward,
    replace_return_docstrings,
)
from ..paligemma import (
    PaliGemmaForConditionalGeneration,
)


@dataclass
class ColPaliForRetrievalOutput(ModelOutput):
    """
    Base class for ColPali embeddings output.

    Args:
        embeddings (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
            The embeddings of the model.
        loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided):
            Language modeling loss (for next-token prediction).
        past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
            Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
            `(batch_size, num_heads, sequence_length, embed_size_per_head)`)

            Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see
            `past_key_values` input) to speed up sequential decoding.
        hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
            Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.

            Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.
        attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
            sequence_length)`.

            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
            heads.
        image_hidden_states (`torch.FloatTensor`, *optional*):
            A `torch.FloatTensor` of size `(batch_size, num_images, sequence_length, hidden_size)`.
            image_hidden_states of the model produced by the vision encoder after projecting last hidden state.
    """

    embeddings: torch.Tensor = None
    loss: Optional[torch.FloatTensor] = None
    past_key_values: Optional[Union[List[torch.FloatTensor], Cache]] = None
    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
    attentions: Optional[Tuple[torch.FloatTensor]] = None
    image_hidden_states: Optional[torch.FloatTensor] = None


@add_start_docstrings(
    """
    ColPali leverages Vision Language Models (VLMs) to construct efficient multi-vector embeddings in the visual space for document retrieval.
    By feeding the ViT output patches from PaliGemma-3B to a linear projection, we create a multi-vector representation of documents. The model
    is trained to maximize the similarity between these document embeddings and the query embeddings, following the ColBERT method.

    Using ColPali removes the need for potentially complex and brittle layout recognition and OCR pipelines with a single model that can take into account
    both the textual and visual content (layout, charts, ...) of a document.

    ColPali was introduced in the following paper: [*ColPali: Efficient Document Retrieval with Vision Language Models*](https://arxiv.org/abs/2407.01449).

    Resources:
    - A blog post detailing ColPali, a vision retrieval model, can be found [here](https://huggingface.co/blog/manu/colpali). 📝
    - The code for training ColPali and for the `colpali-engine` package can be found [here](https://github.com/illuin-tech/colpali). 🌎
    - Cookbooks to fine-tune ColPali (with optional quantization), generate similarity maps, ... can be found [here](https://github.com/tonywu71/colpali-cookbooks). 📚

    Adapted from [`colpali-engine==0.3.0`](https://github.com/illuin-tech/colpali/releases/tag/v0.3.0).
    """
)
class ColPaliForRetrieval(PaliGemmaForConditionalGeneration):
    main_input_name: ClassVar[str] = "doc_input_ids"  # transformers-related

    def __init__(self, config: ColPaliConfig):
        super().__init__(config=config)

        self.embedding_dim = self.config.embedding_dim
        self.custom_text_proj = nn.Linear(self.config.text_config.hidden_size, self.embedding_dim)

        if self.language_model._tied_weights_keys is not None:
            self._tied_weights_keys = [f"model.language_model.{k}" for k in self.language_model._tied_weights_keys]

        self.post_init()

    @add_start_docstrings_to_model_forward(
        """
        Args:
        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
            it.
            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.
            [What are input IDs?](../glossary#input-ids)
        pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, image_size, image_size)):
            The tensors corresponding to the input images. Pixel values can be obtained using
            [`AutoImageProcessor`]. See [`SiglipImageProcessor.__call__`] for details ([]`PaliGemmaProcessor`] uses
            [`SiglipImageProcessor`] for processing images). If none, ColPali will only process text (query embeddings).
        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
            - 1 for tokens that are **not masked**,
            - 0 for tokens that are **masked**.
            [What are attention masks?](../glossary#attention-mask)
            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.
            If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see
            `past_key_values`).
            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
            information on the default strategy.
            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.
        """
    )
    @replace_return_docstrings(output_type=ColPaliForRetrievalOutput, config_class="ColPaliConfig")
    def forward(
        self,
        input_ids: torch.LongTensor = None,
        pixel_values: torch.FloatTensor = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Union[List[torch.FloatTensor], Cache]] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        cache_position: Optional[torch.LongTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        num_logits_to_keep: int = 0,
    ) -> Union[Tuple, ColPaliForRetrievalOutput]:
        r"""
        Returns:
        """
        vlm_outputs = super().forward(
            input_ids=input_ids,
            pixel_values=pixel_values,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            token_type_ids=token_type_ids,
            cache_position=cache_position,
            inputs_embeds=inputs_embeds,
            labels=labels,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=True,
            return_dict=True,
            num_logits_to_keep=num_logits_to_keep,
        )
        last_hidden_states = vlm_outputs.hidden_states[-1]  # (batch_size, sequence_length, hidden_size)
        proj = self.custom_text_proj(last_hidden_states)  # (batch_size, sequence_length, dim)

        # L2 normalization
        embeddings = proj / proj.norm(dim=-1, keepdim=True)  # (batch_size, sequence_length, dim)

        embeddings = embeddings * attention_mask.unsqueeze(-1)  # (batch_size, sequence_length, dim)

        if not return_dict:
            return (embeddings,) + vlm_outputs

        return ColPaliForRetrievalOutput(
            embeddings=embeddings,
            past_key_values=vlm_outputs.past_key_values,
            hidden_states=vlm_outputs.hidden_states,
            attentions=vlm_outputs.attentions,
            image_hidden_states=vlm_outputs.image_hidden_states,
        )

    def resize_token_embeddings(
        self,
        new_num_tokens: Optional[int] = None,
        pad_to_multiple_of=None,
    ) -> nn.Embedding:
        model_embeds = self.language_model.resize_token_embeddings(new_num_tokens, pad_to_multiple_of)

        # Update vocab size
        self.config.text_config.vocab_size = model_embeds.num_embeddings
        self.config.vocab_size = model_embeds.num_embeddings
        self.vocab_size = model_embeds.num_embeddings

        return model_embeds

Expected behavior

Issues

There are 2 problems in the resulting file:

1. There is a missing from ..colpali import ColPaliConfig.
2. The ColPaliForRetrievalOutput is missing.

My initial investigation

After debugging the modular_model_converter.py script part with:

        for file, node in cst_transformers.files.items():
            if node != {}:
                ruffed_code = run_ruff(AUTO_GENERATED_MESSAGE + node.code, True)
                formatted_code = run_ruff(ruffed_code, False)
                output[file] = [formatted_code, ruffed_code]

I observed that the 2 elements mentioned above are already missing from node.code.

Could I get some help on this issue please?
Thanks in advance! 😁

[yonigozlan]

It looks like most of the issues are fixed if we define

class ColPaliForConditionalGeneration(PaliGemmaForConditionalGeneration):

Instead of

class ColPaliForRetrieval(PaliGemmaForConditionalGeneration):

Of course this is not what we want here.
Is this expected with the current state of modular @ArthurZucker ?
If it is, it would be great if we could support task-specific models inheriting from other task-specific (but different task) models. It's a bit weird but I think it can make sense, as the example of ColPali here shows.

[ArthurZucker]

Hey! This will be partly adressed by #33859!
One issue is that if your model name is ColPali, and the original name is PaliGemma, the converter has to guess the renaming patterns (PaliGemmaAttention -> ColPaliAttention) based only on the name!
Will test that, thanks for testing and reporting 🤗