Various fixes for VLMs (#3063)
* fix to not use batch feature indexing * more vlm fixes * use AutoModelForImageTextToText * add example yaml and need num2words for chat template * improve handling of adding image tokens to conversation * add lfm2-vl support * update the lfm readme * fix markdown and add rtol for loss checks * feat: add smolvlm2 processing strat * fix: check for causal-conv1d in lfm models * feat: add docs for lfm2 * feat: add new models and tips to docs * feat: add smolvlm2 docs and remove extra dep * chore: update docs * feat: add video instructions * chore: cleanup * chore: comments * fix: typo * feat: add usage stats * chore: refactor --------- Co-authored-by: NanoCode012 <nano@axolotl.ai>
This commit is contained in:
Wing Lian
GitHub
58
examples/LiquidAI/README.md
Normal file
58
examples/LiquidAI/README.md
Normal file
@@ -0,0 +1,58 @@
|
||||
# Finetune Liquid Foundation Models 2 (LFM2) with Axolotl
|
||||
|
||||
[Liquid Foundation Models 2 (LFM2)](https://huggingface.co/collections/LiquidAI/lfm2-686d721927015b2ad73eaa38) are a family of small, open-weight models from [Liquid AI](https://www.liquid.ai/) focused on quality, speed, and memory efficiency. Liquid AI released text-only [LFM2](https://huggingface.co/collections/LiquidAI/lfm2-686d721927015b2ad73eaa38) and text+vision [LFM2-VL](https://huggingface.co/collections/LiquidAI/lfm2-vl-68963bbc84a610f7638d5ffa) models.
|
||||
|
||||
LFM2 features a new hybrid Liquid architecture with multiplicative gates, short-range convolutions, and grouped query attention, enabling fast training and inference.
|
||||
|
||||
This guide shows how to fine-tune both the LFM2 and LFM2-VL models with Axolotl.
|
||||
|
||||
## Getting Started
|
||||
|
||||
1. Install Axolotl following the [installation guide](https://docs.axolotl.ai/docs/installation.html).
|
||||
|
||||
Here is an example of how to install from pip:
|
||||
```bash
|
||||
# Ensure you have a compatible version of Pytorch installed
|
||||
pip3 install packaging setuptools wheel ninja
|
||||
pip3 install --no-build-isolation 'axolotl[flash-attn]>=0.12.0'
|
||||
```
|
||||
|
||||
2. Run one of the finetuning examples below.
|
||||
|
||||
**LFM2**
|
||||
```bash
|
||||
# FFT SFT (1x48GB @ 25GiB)
|
||||
axolotl train examples/LiquidAI/lfm2-350m-fft.yaml
|
||||
```
|
||||
|
||||
**LFM2-VL**
|
||||
```bash
|
||||
# LoRA SFT (1x48GB @ 2.7GiB)
|
||||
axolotl train examples/LiquidAI/lfm2-vl-lora.yaml
|
||||
```
|
||||
|
||||
### TIPS
|
||||
|
||||
- **Installation Error**: If you encounter `ImportError: ... undefined symbol ...` or `ModuleNotFoundError: No module named 'causal_conv1d_cuda'`, the `causal-conv1d` package may have been installed incorrectly. Try uninstalling it:
|
||||
```bash
|
||||
pip uninstall -y causal-conv1d
|
||||
```
|
||||
|
||||
- **Dataset Loading**: Read more on how to load your own dataset in our [documentation](https://docs.axolotl.ai/docs/dataset_loading.html).
|
||||
- **Dataset Formats**:
|
||||
- For LFM2 models, the dataset format follows the OpenAI Messages format as seen [here](https://docs.axolotl.ai/docs/dataset-formats/conversation.html#chat_template).
|
||||
- For LFM2-VL models, Axolotl follows the multi-content Messages format. See our [Multimodal docs](https://docs.axolotl.ai/docs/multimodal.html#dataset-format) for details.
|
||||
|
||||
## Optimization Guides
|
||||
|
||||
- [Multi-GPU Training](https://docs.axolotl.ai/docs/multi-gpu.html)
|
||||
- [LoRA Optimizations](https://docs.axolotl.ai/docs/lora_optims.html)
|
||||
- [Multi-Node Training](https://docs.axolotl.ai/docs/multi-node.html)
|
||||
|
||||
## Related Resources
|
||||
|
||||
- [LFM2 Blog](https://www.liquid.ai/blog/liquid-foundation-models-v2-our-second-series-of-generative-ai-models)
|
||||
- [LFM2-VL Blog](https://www.liquid.ai/blog/lfm2-vl-efficient-vision-language-models)
|
||||
- [Axolotl Docs](https://docs.axolotl.ai)
|
||||
- [Axolotl GitHub](https://github.com/axolotl-ai-cloud/axolotl)
|
||||
- [Axolotl Discord](https://discord.gg/7m9sfhzaf3)
|
||||
@@ -2,7 +2,6 @@ base_model: LiquidAI/LFM2-350M
|
||||
|
||||
chunked_cross_entropy: true
|
||||
|
||||
chat_template: tokenizer_default
|
||||
eot_tokens:
|
||||
- "<|im_end|>"
|
||||
datasets:
|
||||
58
examples/LiquidAI/lfm2-vl-lora.yaml
Normal file
58
examples/LiquidAI/lfm2-vl-lora.yaml
Normal file
@@ -0,0 +1,58 @@
|
||||
base_model: LiquidAI/LFM2-VL-450M
|
||||
trust_remote_code: true
|
||||
model_type: AutoModelForImageTextToText
|
||||
processor_type: AutoProcessor
|
||||
|
||||
# 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%]
|
||||
|
||||
dataset_prepared_path: last_run_prepared
|
||||
val_set_size: 0.0
|
||||
output_dir: ./outputs/out
|
||||
|
||||
adapter: lora
|
||||
lora_model_dir:
|
||||
|
||||
sequence_len: 8192
|
||||
pad_to_sequence_len: false
|
||||
|
||||
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: 1
|
||||
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
|
||||
@@ -1,7 +0,0 @@
|
||||
# Liquid Foundation Models 2
|
||||
|
||||
LFM2 support in transformers exists in the main branch, but is not yet included in the transformers release.
|
||||
|
||||
```bash
|
||||
pip install --upgrade --no-deps --force-reinstall git+https://github.com/huggingface/transformers.git
|
||||
```
|
||||
49
examples/smolvlm2/README.md
Normal file
49
examples/smolvlm2/README.md
Normal file
@@ -0,0 +1,49 @@
|
||||
# Finetune SmolVLM2 with Axolotl
|
||||
|
||||
[SmolVLM2](https://huggingface.co/collections/HuggingFaceTB/smolvlm2-smallest-video-lm-ever-67ab6b5e84bf8aaa60cb17c7) are a family of lightweight, open-source multimodal models from HuggingFace designed to analyze and understand video, image, and text content.
|
||||
|
||||
These models are built for efficiency, making them well-suited for on-device applications where computational resources are limited. Models are available in multiple sizes, including 2.2B, 500M, and 256M.
|
||||
|
||||
This guide shows how to fine-tune SmolVLM2 models with Axolotl.
|
||||
|
||||
## Getting Started
|
||||
|
||||
1. Install Axolotl following the [installation guide](https://docs.axolotl.ai/docs/installation.html).
|
||||
|
||||
Here is an example of how to install from pip:
|
||||
```bash
|
||||
# Ensure you have a compatible version of Pytorch installed
|
||||
pip3 install packaging setuptools wheel ninja
|
||||
pip3 install --no-build-isolation 'axolotl[flash-attn]>=0.12.0'
|
||||
```
|
||||
|
||||
2. Install an extra dependency:
|
||||
|
||||
```bash
|
||||
pip3 install num2words==0.5.14
|
||||
```
|
||||
|
||||
3. Run the finetuning example:
|
||||
|
||||
```bash
|
||||
# LoRA SFT (1x48GB @ 6.8GiB)
|
||||
axolotl train examples/smolvlm2/smolvlm2-2B-lora.yaml
|
||||
```
|
||||
|
||||
## TIPS
|
||||
|
||||
- **Dataset Format**: For video finetuning, your dataset must be compatible with the multi-content Messages format. For more details, see our documentation on [Multimodal Formats](https://docs.axolotl.ai/docs/multimodal.html#dataset-format).
|
||||
- **Dataset Loading**: Read more on how to prepare and load your own datasets in our [documentation](https://docs.axolotl.ai/docs/dataset_loading.html).
|
||||
|
||||
## Optimization Guides
|
||||
|
||||
- [Multi-GPU Training](https://docs.axolotl.ai/docs/multi-gpu.html)
|
||||
- [LoRA Optimizations](https://docs.axolotl.ai/docs/lora_optims.html)
|
||||
- [Multi-Node Training](https://docs.axolotl.ai/docs/multi-node.html)
|
||||
|
||||
## Related Resources
|
||||
|
||||
- [SmolVLM2 Blog](https://huggingface.co/blog/smolvlm2)
|
||||
- [Axolotl Docs](https://docs.axolotl.ai)
|
||||
- [Axolotl GitHub](https://github.com/axolotl-ai-cloud/axolotl)
|
||||
- [Axolotl Discord](https://discord.gg/7m9sfhzaf3)
|
||||
56
examples/smolvlm2/smolvlm2-2B-lora.yaml
Normal file
56
examples/smolvlm2/smolvlm2-2B-lora.yaml
Normal file
@@ -0,0 +1,56 @@
|
||||
base_model: HuggingFaceTB/SmolVLM2-2.2B-Instruct
|
||||
trust_remote_code: true
|
||||
processor_type: AutoProcessor
|
||||
|
||||
# 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%]
|
||||
dataset_prepared_path: last_run_prepared
|
||||
val_set_size: 0.0
|
||||
output_dir: ./outputs/out
|
||||
|
||||
adapter: lora
|
||||
lora_model_dir:
|
||||
|
||||
sequence_len: 8192
|
||||
pad_to_sequence_len: false
|
||||
|
||||
lora_r: 32
|
||||
lora_alpha: 16
|
||||
lora_dropout: 0.05
|
||||
lora_target_modules: 'model.text_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: 1
|
||||
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
|
||||
Reference in New Issue
Block a user