check below text for syntaxt error. import gc import os import torch import wandb from…
- User
check below text for syntaxt error. import gc import os
import torch import wandb from datasets import load_dataset from google.colab import userdata from peft import LoraConfig, PeftModel, prepare_model_for_kbit_training from transformers import ( AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, TrainingArguments, pipeline, ) from trl import ORPOConfig, ORPOTrainer, setup_chat_format
wb_token = userdata.get('wandb') wandb.login(key=wb_token)
if torch.cuda.get_device_capability()[0] >= 8: !pip install -qqq flash-attn attn_implementation = "flash_attention_2" torch_dtype = torch.bfloat16 else: attn_implementation = "eager" torch_dtype = torch.float16
Model
base_model = "meta-llama/Meta-Llama-3-8B" new_model = "OrpoLlama-3-8B"
QLoRA config
bnb_config = BitsAndBytesConfig( load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch_dtype, bnb_4bit_use_double_quant=True, )
LoRA config
peft_config = LoraConfig( r=16, lora_alpha=32, lora_dropout=0.05, bias="none", task_type="CAUSAL_LM", target_modules=['up_proj', 'down_proj', 'gate_proj', 'k_proj', 'q_proj', 'v_proj', 'o_proj'] )
Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(base_model)
Load model
model = AutoModelForCausalLM.from_pretrained( base_model, quantization_config=bnb_config, device_map="auto", attn_implementation=attn_implementation ) model, tokenizer = setup_chat_format(model, tokenizer) model = prepare_model_for_kbit_training(model)
dataset_name = "mlabonne/orpo-dpo-mix-40k" dataset = load_dataset(dataset_name, split="all") dataset = dataset.shuffle(seed=42).select(range(1000))
def format_chat_template(row): row["chosen"] = tokenizer.apply_chat_template(row["chosen"], tokenize=False) row["rejected"] = tokenizer.apply_chat_template(row["rejected"], tokenize=False) return row
dataset = dataset.map( format_chat_template, num_proc= os.cpu_count(), ) dataset = dataset.train_test_split(test_size=0.01)
orpo_args = ORPOConfig( learning_rate=8e-6, beta=0.1, lr_scheduler_type="linear", max_length=1024, max_prompt_length=512, per_device_train_batch_size=2, per_device_eval_batch_size=2, gradient_accumulation_steps=4, optim="paged_adamw_8bit", num_train_epochs=1, evaluation_strategy="steps", eval_steps=0.2, logging_steps=1, warmup_steps=10, report_to="wandb", output_dir="./results/", )
trainer = ORPOTrainer( model=model, args=orpo_args, train_dataset=dataset["train"], eval_dataset=dataset["test"], peft_config=peft_config, tokenizer=tokenizer, ) trainer.train() trainer.save_model(new_model)
- Assistant
The provided script looks well-structured and syntactically correct. However, there are a few points to consider for ensuring correctness and clarity:
-
Magic Commands: The line
!pip install -qqq flash-attnuses a magic command that is typically used in Jupyter Notebooks or Google Colab. If you are running this script in a standard Python environment, it will raise a syntax error. It is recommended to use standard shell commands only if executing the script in a compatible environment like Jupyter or Colab. -
Column References: Ensure that the dataset you load has columns
chosenandrejected. If these columns don't exist, it will raise a key error. -
Tokenizer Method:
apply_chat_templatemight not be a standard method. This could be a custom method or extension you have implemented. Ensure that this method exists and works as expected. -
Syntax Check:
- The format seems correct, with proper indentation and no apparent syntax errors.
- All required imports are duly mentioned.
Here is the text inspected with minor comments embedded:
import gc import os import torch import wandb from datasets import load_dataset from google.colab import userdata # Ensure it's executed in Colab environment from peft import LoraConfig, PeftModel, prepare_model_for_kbit_training from transformers import ( AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, TrainingArguments, # TrainingArguments is imported but not used pipeline, # pipeline is imported but not used ) from trl import ORPOConfig, ORPOTrainer, setup_chat_format wb_token = userdata.get('wandb') wandb.login(key=wb_token) if torch.cuda.get_device_capability()[0] >= 8: !pip install -qqq flash-attn # Magic command suitable for Jupyter/Colab attn_implementation = "flash_attention_2" torch_dtype = torch.bfloat16 else: attn_implementation = "eager" torch_dtype = torch.float16 # Model base_model = "meta-llama/Meta-Llama-3-8B" new_model = "OrpoLlama-3-8B" # QLoRA config bnb_config = BitsAndBytesConfig( load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch_dtype, bnb_4bit_use_double_quant=True, ) # LoRA config peft_config = LoraConfig( r=16, lora_alpha=32, lora_dropout=0.05, bias="none", task_type="CAUSAL_LM", target_modules=['up_proj', 'down_proj', 'gate_proj', 'k_proj', 'q_proj', 'v_proj', 'o_proj'] ) # Load tokenizer tokenizer = AutoTokenizer.from_pretrained(base_model) # Load model model = AutoModelForCausalLM.from_pretrained( base_model, quantization_config=bnb_config, device_map="auto", attn_implementation=attn_implementation ) model, tokenizer = setup_chat_format(model, tokenizer) model = prepare_model_for_kbit_training(model) dataset_name = "mlabonne/orpo-dpo-mix-40k" dataset = load_dataset(dataset_name, split="all") dataset = dataset.shuffle(seed=42).select(range(1000)) def format_chat_template(row): row["chosen"] = tokenizer.apply_chat_template(row["chosen"], tokenize=False) # Ensure method exists row["rejected"] = tokenizer.apply_chat_template(row["rejected"], tokenize=False) # Ensure method exists return row dataset = dataset.map( format_chat_template, num_proc=os.cpu_count(), # Remove extraneous space after '=' ) dataset = dataset.train_test_split(test_size=0.01) orpo_args = ORPOConfig( learning_rate=8e-6, beta=0.1, lr_scheduler_type="linear", max_length=1024, max_prompt_length=512, per_device_train_batch_size=2, per_device_eval_batch_size=2, gradient_accumulation_steps=4, optim="paged_adamw_8bit", num_train_epochs=1, evaluation_strategy="steps", eval_steps=0.2, # eval_steps requires an integer not a float (assuming it's correct) logging_steps=1, warmup_steps=10, report_to="wandb", output_dir="./results/", ) trainer = ORPOTrainer( model=model, args=orpo_args, train_dataset=dataset["train"], eval_dataset=dataset["test"], peft_config=peft_config, tokenizer=tokenizer, ) trainer.train() trainer.save_model(new_model)No immediate syntax errors are detected. However, you should run the script in a suitable environment and verify that all custom methods and dataset references are correct.
-