diff --git a/examples/sentiment/scripts/gpt-neox-20b_peft/clm_finetune_peft_imdb.py b/examples/sentiment/scripts/gpt-neox-20b_peft/clm_finetune_peft_imdb.py index b78111605c..7ce3cc400f 100644 --- a/examples/sentiment/scripts/gpt-neox-20b_peft/clm_finetune_peft_imdb.py +++ b/examples/sentiment/scripts/gpt-neox-20b_peft/clm_finetune_peft_imdb.py @@ -59,9 +59,7 @@ class DataTrainingArguments: if "gpt-neox" in model_args.model_name_or_path: - model = prepare_model_for_int8_training( - model, output_embedding_layer_name="embed_out", layer_norm_names=["layer_norm", "layernorm"] - ) + model = prepare_model_for_int8_training(model, output_embedding_layer_name="embed_out") else: model = prepare_model_for_int8_training(model) diff --git a/examples/sentiment/scripts/gpt2-sentiment_peft.py b/examples/sentiment/scripts/gpt2-sentiment_peft.py index 98274864b0..fbb2b23947 100644 --- a/examples/sentiment/scripts/gpt2-sentiment_peft.py +++ b/examples/sentiment/scripts/gpt2-sentiment_peft.py @@ -152,7 +152,6 @@ def collator(data): config.model_name, load_in_8bit=True, peft_config=lora_config, - layer_norm_names=[], ) tokenizer = AutoTokenizer.from_pretrained(config.model_name) diff --git a/examples/stack_llama/scripts/rl_training.py b/examples/stack_llama/scripts/rl_training.py index f38c654f73..0d8c9b1ce8 100644 --- a/examples/stack_llama/scripts/rl_training.py +++ b/examples/stack_llama/scripts/rl_training.py @@ -183,7 +183,6 @@ def collator(data): load_in_8bit=True, device_map={"": current_device}, peft_config=lora_config, - layer_norm_names=[], ) optimizer = None diff --git a/requirements.txt b/requirements.txt index 8c7d9c6c25..b4c9eed937 100644 --- a/requirements.txt +++ b/requirements.txt @@ -3,3 +3,4 @@ torch>=1.4.0 tqdm transformers accelerate +peft>=0.3.0