diff --git a/src/llama_recipes/finetuning.py b/src/llama_recipes/finetuning.py index e1d702e2005b557650e63f5d6c424a03bf686daf..2a86234c30857d58e562736813f229d8171f39f5 100644 --- a/src/llama_recipes/finetuning.py +++ b/src/llama_recipes/finetuning.py @@ -237,7 +237,8 @@ def main(**kwargs): if not train_config.use_peft and train_config.freeze_LLM_only and config.model_type == "mllama": freeze_LLM_only(model) - + + print_model_size(model, train_config, rank if train_config.enable_fsdp else 0) mixed_precision_policy, wrapping_policy = get_policies(fsdp_config, rank) # Create the FSDP wrapper for MllamaSelfAttentionDecoderLayer,MllamaSelfAttentionDecoderLayer,MllamaVisionEncoderLayer in vision models @@ -306,8 +307,6 @@ def main(**kwargs): dataset_processer = processor else: dataset_processer = tokenizer - - print_model_size(model, train_config, rank if train_config.enable_fsdp else 0) # Load and preprocess the dataset for training and validation