axolotl-ai-cloud · NanoCode012 · Dec 25, 2025 · Sep 8, 2025 · Sep 10, 2025 · Dec 24, 2025
diff --git a/README.md b/README.md
@@ -29,7 +29,7 @@
 
 ## 🎉 Latest Updates
 
-- 2025/12: Axolotl now includes support for [Kimi-Linear](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/kimi-linear), [Olmo3](https://github.com/axolotl-ai-cloud/axolotl/blob/main/examples/olmo3), [Trinity](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/trinity), and [Ministral3](https://github.com/axolotl-ai-cloud/axolotl/blob/main/examples/ministral3).
+- 2025/12: Axolotl now includes support for [Kimi-Linear](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/kimi-linear), [InternVL 3.5](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/internvl3_5), [Olmo3](https://github.com/axolotl-ai-cloud/axolotl/blob/main/examples/olmo3), [Trinity](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/trinity), and [Ministral3](https://github.com/axolotl-ai-cloud/axolotl/blob/main/examples/ministral3).
 - 2025/10: New model support has been added in Axolotl for: [Qwen3 Next](https://github.com/axolotl-ai-cloud/axolotl/blob/main/examples/qwen3-next), [Qwen2.5-vl, Qwen3-vl](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/qwen2_5-vl), [Qwen3, Qwen3MoE](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/qwen3), [Granite 4](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/granite4), [HunYuan](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/hunyuan), [Magistral 2509](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/magistral#vision), [Apertus](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/apertus), and [Seed-OSS](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/seed-oss).
 - 2025/09: Axolotl now has text diffusion training. Read more [here](https://github.com/axolotl-ai-cloud/axolotl/tree/main/src/axolotl/integrations/diffusion).
 - 2025/08: QAT has been updated to include NVFP4 support. See [PR](https://github.com/axolotl-ai-cloud/axolotl/pull/3107).

diff --git a/docs/multimodal.qmd b/docs/multimodal.qmd
@@ -21,6 +21,7 @@ format:
 - [Qwen2.5-VL](#sec-qwen25-vl)
 - [SmolVLM2](#sec-smolvlm2)
 - [LFM2-VL](#sec-lfm2-vl)
+- [Intern-VL](#sec-intern-vl)
 
 ## Usage
 
@@ -202,6 +203,16 @@ Please uninstall `causal-conv1d` via `pip3 uninstall -y causal-conv1d`
 base_model: LiquidAI/LFM2-VL-450M
 ```
 
+### Intern-VL {#sec-intern-vl}
+
+::: {.callout-tip}
+Please make sure to install `timm` via `pip3 install timm==1.0.19`
+:::
+
+```yaml
+base_model: OpenGVLab/InternVL3_5-8B
+```
+
 ## Dataset Format
 
 For multi-modal datasets, we adopt an extended `chat_template` format similar to OpenAI's Message format.

diff --git a/examples/colab-notebooks/colab-axolotl-example.ipynb b/examples/colab-notebooks/colab-axolotl-example.ipynb
@@ -40,7 +40,7 @@
     "%%capture\n",
     "# This step can take ~5-10 minutes to install dependencies\n",
     "!pip install --no-build-isolation axolotl[flash-attn]>=0.9.1\n",
-    "!pip install \"cut-cross-entropy[transformers] @ git+https://github.com/axolotl-ai-cloud/ml-cross-entropy.git@242b245\""
+    "!pip install \"cut-cross-entropy[transformers] @ git+https://github.com/axolotl-ai-cloud/ml-cross-entropy.git@318b7e2\""
    ]
   },
   {

diff --git a/examples/internvl3_5/README.md b/examples/internvl3_5/README.md
@@ -0,0 +1,43 @@
+# Finetune OpenGV's InternVL with Axolotl
+
+[InternVL 3.5](https://huggingface.co/OpenGVLab/InternVL3_5-8B-HF) is a family of powerful vision-language models supporting dynamic resolution and multi-image understanding by OpenGV. It features a ViT-style vision encoder and strong language model backbone for tasks like visual question answering, OCR, and scene text understanding.
+
+This guide shows how to fine-tune it with Axolotl.
+
+## Getting started
+
+1. Install Axolotl following the [installation guide](https://docs.axolotl.ai/docs/installation.html).
+
+2. Install `timm` for vision model support:
+
+    ```bash
+    pip install timm==1.0.19
+    ```
+
+3. Install [Cut Cross Entropy](https://docs.axolotl.ai/docs/custom_integrations.html#cut-cross-entropy) to reduce training VRAM usage.
+
+4. Run the finetuning example:
+
+    ```bash
+    axolotl train examples/internvl3_5/internvl3_5-8b-qlora.yml
+    ```
+
+This config uses about 8.21 GiB VRAM. Let us know how it goes. Happy finetuning! 🚀
+
+### Tips
+
+- You can run a full finetuning by removing the `adapter: qlora` and `load_in_4bit: true` from the config.
+- Read more on how to load your own dataset at [docs](https://docs.axolotl.ai/docs/dataset_loading.html).
+- The dataset format follows the multi-modal format as seen [here](https://docs.axolotl.ai/docs/multimodal.html#dataset-format).
+
+## Optimization Guides
+
+Please check the [Optimizations doc](https://docs.axolotl.ai/docs/optimizations.html).
+
+## Related Resources
+
+- [InternVL Paper](https://huggingface.co/papers/2508.18265)
+- [Axolotl Docs](https://docs.axolotl.ai)
+- [Axolotl Website](https://axolotl.ai)
+- [Axolotl GitHub](https://github.com/axolotl-ai-cloud/axolotl)
+- [Axolotl Discord](https://discord.gg/7m9sfhzaf3)
diff --git a/examples/internvl3_5/internvl3_5-8b-qlora.yml b/examples/internvl3_5/internvl3_5-8b-qlora.yml
@@ -0,0 +1,61 @@
+base_model: OpenGVLab/InternVL3_5-8B-HF
+processor_type: AutoProcessor
+
+plugins:
+  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
+
+load_in_4bit: true
+
+# these 3 lines are needed for now to handle vision chat templates w images
+skip_prepare_dataset: true
+remove_unused_columns: false
+sample_packing: false
+
+datasets:
+  - path: HuggingFaceH4/llava-instruct-mix-vsft
+    type: chat_template
+    split: train[:1%]
+    field_messages: messages
+
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.01
+output_dir: ./outputs/out
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 2048
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+bf16: true
+fp16:
+tf32: true
+
+gradient_checkpointing: true
+logging_steps: 1
+flash_attention: true
+eager_attention:
+
+warmup_ratio: 0.1
+evals_per_epoch: 1
+saves_per_epoch: 1
+weight_decay: 0.0
+
+# save_first_step: true  # uncomment this to validate checkpoint saving works with your config
diff --git a/scripts/cutcrossentropy_install.py b/scripts/cutcrossentropy_install.py
@@ -29,5 +29,5 @@
 
 print(
     UNINSTALL_PREFIX
-    + f'{UV_PREFIX}pip install "cut-cross-entropy[transformers] @ git+https://github.com/axolotl-ai-cloud/ml-cross-entropy.git@242b245"'
+    + f'{UV_PREFIX}pip install "cut-cross-entropy[transformers] @ git+https://github.com/axolotl-ai-cloud/ml-cross-entropy.git@318b7e2"'
 )
diff --git a/src/axolotl/integrations/cut_cross_entropy/README.md b/src/axolotl/integrations/cut_cross_entropy/README.md
@@ -19,7 +19,7 @@ python scripts/cutcrossentropy_install.py | sh
 
 - If you are installing from pip
 ```bash
-pip3 uninstall -y cut-cross-entropy && pip3 install "cut-cross-entropy[transformers] @ git+https://github.com/axolotl-ai-cloud/ml-cross-entropy.git@242b245"
+pip3 uninstall -y cut-cross-entropy && pip3 install "cut-cross-entropy[transformers] @ git+https://github.com/axolotl-ai-cloud/ml-cross-entropy.git@318b7e2"
 ```
 
 ## Usage
@@ -54,6 +54,7 @@ plugins:
 - granitemoehybrid
 - hunyuan_v1_dense
 - hunyuan_v1_moe
+- internvl
 - kimi_linear
 - lfm2
 - lfm2_moe

diff --git a/src/axolotl/integrations/cut_cross_entropy/__init__.py b/src/axolotl/integrations/cut_cross_entropy/__init__.py
@@ -35,7 +35,7 @@
 
 _CCE_INSTALL_MESSAGE = (
     "Please install Axolotl's fork of cut_cross_entropy with transformers support using "
-    '`pip install "cut-cross-entropy[transformers] @ git+https://github.com/axolotl-ai-cloud/ml-cross-entropy.git@242b245"`'
+    '`pip install "cut-cross-entropy[transformers] @ git+https://github.com/axolotl-ai-cloud/ml-cross-entropy.git@318b7e2"`'
 )
 
 

diff --git a/src/axolotl/loaders/utils.py b/src/axolotl/loaders/utils.py
@@ -79,7 +79,11 @@ def check_model_config(cfg: DictDefault, model_config: PretrainedConfig):
             and hasattr(model_config, "vision_config")
             and hasattr(model_config.vision_config, "image_size")
         ):
-            cfg.image_size = model_config.vision_config.image_size
+            image_size = model_config.vision_config.image_size
+            if isinstance(image_size, list):
+                cfg.image_size = tuple(image_size)
+            else:
+                cfg.image_size = image_size
             LOG.debug(f"Loaded image size: {cfg.image_size} from model config")
 
     quant_config_exists = (

diff --git a/src/axolotl/processing_strategies.py b/src/axolotl/processing_strategies.py
@@ -8,6 +8,7 @@
 from torch import Tensor, zeros_like
 from transformers import ProcessorMixin
 from transformers.image_utils import load_image
+from transformers.models.internvl import InternVLProcessor
 from transformers.models.smolvlm import SmolVLMProcessor
 from transformers.models.voxtral import VoxtralProcessor
 
@@ -454,6 +455,37 @@ def process_labels(self, input_ids):
         return labels
 
 
+class InternVLProcessingStrategy(ProcessingStrategy):
+    """Processing Strategy class for InternVL"""
+
+    def __init__(
+        self,
+        processor: ProcessorMixin,
+        chat_template: Optional[str] = None,
+        image_size: int | tuple[int, int] | None = None,
+        image_resize_algorithm: Resampling | None = None,
+    ):
+        super().__init__(processor, chat_template, image_size, image_resize_algorithm)
+
+        if not hasattr(processor, "image_ids"):
+            raise ValueError("'image_ids' missing from InternVL Processor.")
+
+        self.image_token_ids = processor.image_ids
+
+    def process_labels(self, input_ids):
+        labels = input_ids.clone()
+
+        labels[labels == self.processor.tokenizer.pad_token_id] = -100
+
+        for ids in self.image_token_ids:
+            labels[labels == ids] = -100
+
+        # Note: Check if need to mask 'video_token' as it gets converted to
+        # image patches during media processing
+
+        return labels
+
+
 def get_processing_strategy(
     processor: ProcessorMixin,
     chat_template,
@@ -501,6 +533,11 @@ def get_processing_strategy(
             **processing_kwargs,
         )
 
+    if isinstance(processor, InternVLProcessor):
+        return InternVLProcessingStrategy(
+            **processing_kwargs,
+        )
+
     # llama3_2_vision, llama4, llava
     # mistral_v7_tekken, pixtral, lfm2vl
     return ProcessingStrategy(