update README

recipes/quickstart/finetuning/README.md

@@ -54,6 +54,7 @@ It lets us specify the training settings for everything from `model_name` to `da
     output_dir: str = "PATH/to/save/PEFT/model"
     freeze_layers: bool = False
     num_freeze_layers: int = 1
+    freeze_LLM_only: bool = False # Freeze self-attention layers in the language_model. Vision model, multi_modal_projector, cross-attention will be fine-tuned
     quantization: str = None
     one_gpu: bool = False
     save_model: bool = True

recipes/quickstart/finetuning/finetune_vision_model.md

@@ -18,6 +18,12 @@ For **LoRA finetuning with FSDP**, we can run the following code:
 ```bash
   torchrun --nnodes 1 --nproc_per_node 4  recipes/quickstart/finetuning/finetuning.py --enable_fsdp --lr 1e-5  --num_epochs 3 --batch_size_training 2 --model_name meta-llama/Llama-3.2-11B-Vision-Instruct --dist_checkpoint_root_folder ./finetuned_model --dist_checkpoint_folder fine-tuned  --use_fast_kernels --dataset "custom_dataset" --custom_dataset.test_split "test" --custom_dataset.file "recipes/quickstart/finetuning/datasets/ocrvqa_dataset.py"  --run_validation True --batching_strategy padding  --use_peft --peft_method lora
 ```
+
+For **finetuning with LLM freeze using FSDP**, we can run the following code:
+
+```bash
+  torchrun --nnodes 1 --nproc_per_node 4  recipes/quickstart/finetuning/finetuning.py --enable_fsdp --lr 1e-5  --num_epochs 3 --batch_size_training 2 --model_name meta-llama/Llama-3.2-11B-Vision-Instruct --dist_checkpoint_root_folder ./finetuned_model --dist_checkpoint_folder fine-tuned  --use_fast_kernels --dataset "custom_dataset" --custom_dataset.test_split "test" --custom_dataset.file "recipes/quickstart/finetuning/datasets/ocrvqa_dataset.py"  --run_validation True --batching_strategy padding --freeze_LLM_only True
+```
 **Note**: `--batching_strategy padding` is needed as the vision model will not work with `packing` method.
 
 For more details about the finetuning configurations, please read the [finetuning readme](./README.md).