fix: gemma3 loss in forward pass (#2473) [skip ci]

* fix: gemma3 loss in forward pass

* fix: lint

* fix: move patch before plugins

* Update src/axolotl/monkeypatch/gemma3.py

Co-authored-by: salman <salman.mohammadi@outlook.com>

---------

Co-authored-by: Wing Lian <wing.lian@gmail.com>
Co-authored-by: salman <salman.mohammadi@outlook.com>

src/axolotl/monkeypatch/gemma3.py

@@ -0,0 +1,238 @@
+"""Monkeypatch for gemma3 conditional generation forward to fix loss exploding"""
+
+# pylint: disable=duplicate-code
+
+from typing import Optional, Tuple, Union
+
+import torch
+from transformers.cache_utils import Cache
+from transformers.models.gemma3.modeling_gemma3 import (
+    _CONFIG_FOR_DOC,
+    GEMMA3_INPUTS_DOCSTRING,
+    Gemma3CausalLMOutputWithPast,
+    logger,
+)
+from transformers.utils import (
+    add_start_docstrings_to_model_forward,
+    is_torchdynamo_compiling,
+    replace_return_docstrings,
+)
+from transformers.utils.deprecation import deprecate_kwarg
+
+
+@deprecate_kwarg("num_logits_to_keep", version="4.50", new_name="logits_to_keep")
+@add_start_docstrings_to_model_forward(GEMMA3_INPUTS_DOCSTRING)
+@replace_return_docstrings(
+    output_type=Gemma3CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC
+)
+def new_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,
+    logits_to_keep: Union[int, torch.Tensor] = 0,
+    **lm_kwargs,
+) -> Union[Tuple, Gemma3CausalLMOutputWithPast]:
+    r"""
+        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+            config.text_config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+            (masked), the loss is only computed for the tokens with labels in `[0, ..., config.text_config.vocab_size]`.
+
+        logits_to_keep (`int` or `torch.Tensor`, *optional*):
+            If an `int`, compute logits for the last `logits_to_keep` tokens. If `0`, calculate logits for all
+            `input_ids` (special case). Only last token logits are needed for generation, and calculating them only for that
+            token can save memory, which becomes pretty significant for long sequences or large vocabulary size.
+            If a `torch.Tensor`, must be 1D corresponding to the indices to keep in the sequence length dimension.
+            This is useful when using packed tensor format (single dimension for batch and sequence length).
+
+    Returns:
+
+    Example:
+
+    ```python
+    >>> from PIL import Image
+    >>> import requests
+    >>> from transformers import AutoProcessor, Gemma3ForConditionalGeneration
+
+    >>> model = Gemma3ForConditionalGeneration.from_pretrained("google/Gemma3-test-224px-hf")
+    >>> processor = AutoProcessor.from_pretrained("google/Gemma3-test-224px-hf")
+
+    >>> prompt = "answer en Where is the cow standing?"
+    >>> url = "https://huggingface.co/gv-hf/Gemma3-test-224px-hf/resolve/main/cow_beach_1.png"
+    >>> image = Image.open(requests.get(url, stream=True).raw)
+
+    >>> inputs = processor(images=image, text=prompt,  return_tensors="pt")
+
+    >>> # Generate
+    >>> generate_ids = model.generate(**inputs, max_length=30)
+    >>> processor.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+    "answer en Where is the cow standing?\nbeach"
+    ```"""
+
+    if (input_ids is None) ^ (inputs_embeds is not None):
+        raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
+
+    output_attentions = (
+        output_attentions
+        if output_attentions is not None
+        else self.config.output_attentions
+    )
+    output_hidden_states = (
+        output_hidden_states
+        if output_hidden_states is not None
+        else self.config.output_hidden_states
+    )
+    return_dict = (
+        return_dict if return_dict is not None else self.config.use_return_dict
+    )
+
+    is_training = token_type_ids is not None and labels is not None
+
+    # Replace image id with PAD if the image token is OOV, to avoid index-errors
+    if input_ids is not None and self.config.image_token_index >= self.vocab_size:
+        special_image_mask = input_ids == self.config.image_token_index
+        llm_input_ids = input_ids.clone()
+        llm_input_ids[special_image_mask] = 0
+    else:
+        llm_input_ids = input_ids
+
+    if inputs_embeds is None:
+        inputs_embeds = self.get_input_embeddings()(llm_input_ids)
+
+    if cache_position is None:
+        past_seen_tokens = (
+            past_key_values.get_seq_length() if past_key_values is not None else 0
+        )
+        cache_position = torch.arange(
+            past_seen_tokens,
+            past_seen_tokens + inputs_embeds.shape[1],
+            device=inputs_embeds.device,
+        )
+
+    # Merge text and images
+    if pixel_values is not None:
+        image_features = self.get_image_features(pixel_values)
+
+        if input_ids is None:
+            special_image_mask = inputs_embeds == self.get_input_embeddings()(
+                torch.tensor(
+                    self.config.image_token_index,
+                    dtype=torch.long,
+                    device=inputs_embeds.device,
+                )
+            )
+        else:
+            special_image_mask = (input_ids == self.config.image_token_index).unsqueeze(
+                -1
+            )
+            special_image_mask = special_image_mask.expand_as(inputs_embeds).to(
+                inputs_embeds.device
+            )
+
+        if (
+            not is_torchdynamo_compiling()
+            and inputs_embeds[special_image_mask].numel() != image_features.numel()
+        ):
+            image_tokens_in_text = (special_image_mask).sum(dim=1).sum(dim=0)[0]
+            raise ValueError(
+                f"Number of images does not match number of special image tokens in the input text. "
+                f"Got {image_tokens_in_text} image tokens in the text but {image_features.shape[0] * image_features.shape[1]} "
+                "tokens from image embeddings."
+            )
+        image_features = image_features.to(inputs_embeds.device, inputs_embeds.dtype)
+        inputs_embeds = inputs_embeds.masked_scatter(special_image_mask, image_features)
+
+    # mask out pad-token-ids in labels for BC
+    if labels is not None and self.pad_token_id in labels:
+        logger.warning_once(
+            "`labels` contains `pad_token_id` which will be masked with `config.ignore_index`. "
+            "You have to mask out `pad_token_id` when preparing `labels`, this behavior will be removed in v.4.46.",
+        )
+        labels = torch.where(
+            input_ids == self.pad_token_id, self.config.ignore_index, labels
+        )
+
+    causal_mask = self._update_causal_mask(  # pylint: disable=protected-access
+        attention_mask,
+        token_type_ids,
+        past_key_values,
+        cache_position,
+        inputs_embeds,
+        is_training,
+    )
+    outputs = self.language_model(
+        attention_mask=causal_mask,
+        position_ids=position_ids,
+        past_key_values=past_key_values,
+        inputs_embeds=inputs_embeds,
+        use_cache=use_cache,
+        output_attentions=output_attentions,
+        output_hidden_states=output_hidden_states,
+        return_dict=return_dict,
+        cache_position=cache_position,
+        logits_to_keep=logits_to_keep,
+        **lm_kwargs,
+    )
+
+    logits = outputs[0]
+    loss = None
+    if labels is not None:
+        if attention_mask is not None:
+            # Get the shifted attention mask
+            shift_attention_mask = attention_mask[:, -logits.shape[1] + 1 :].to(
+                logits.device
+            )  # +1 for shift
+
+            # Filter logits and labels based on attention mask
+            valid_indices = shift_attention_mask != 0
+            filtered_logits = logits[..., :-1, :][valid_indices]
+            filtered_labels = labels[..., 1:][valid_indices.to(labels.device)]
+
+            # TODO: do we need to handle num_items_in_batch given we filter the logits and labels?
+
+            loss = self.loss_function(
+                logits=filtered_logits,
+                labels=None,  # we pass shift_labels
+                shift_labels=filtered_labels,
+                vocab_size=self.config.text_config.vocab_size,
+                **lm_kwargs,
+            )
+        else:
+            # Standard case without filtering
+            loss = self.loss_function(
+                logits=logits,
+                labels=labels,
+                vocab_size=self.config.text_config.vocab_size,
+                **lm_kwargs,
+            )
+    if not return_dict:
+        output = (logits,) + outputs[1:]
+        return (loss,) + output if loss is not None else output
+
+    return Gemma3CausalLMOutputWithPast(
+        loss=loss,
+        logits=logits,
+        past_key_values=outputs.past_key_values,
+        hidden_states=outputs.hidden_states,
+        attentions=outputs.attentions,
+        image_hidden_states=image_features if pixel_values is not None else None,
+    )
+
+
+def patch_gemma3conditionalgeneration_forward():
+    from transformers.models.gemma3.modeling_gemma3 import (
+        Gemma3ForConditionalGeneration,
+    )
+
+    Gemma3ForConditionalGeneration.forward = new_forward

src/axolotl/utils/models.py

@@ -535,6 +535,15 @@ class ModelLoader:
         self.auto_model_loader = AutoModelForCausalLM  # pylint: disable=invalid-name
 
     def apply_patches(self) -> None:
+        # patch gemma3 conditional generation forward before loading plugins
+        # as it could be overridden by plugins
+        if self.cfg.model_config_type == "gemma3":
+            from axolotl.monkeypatch.gemma3 import (
+                patch_gemma3conditionalgeneration_forward,
+            )
+
+            patch_gemma3conditionalgeneration_forward()
+
         # load any patches from plugins
         from axolotl.integrations.base import PluginManager