RLLMv3-hero / train.py
migueldeguzmandev's picture
Upload 11 files
8476730 verified
raw
history blame
4.2 kB
import os
os.environ['KMP_DUPLICATE_LIB_OK'] = 'TRUE'
import sys
import torch
from transformers import GPT2Tokenizer, GPT2LMHeadModel, TextDataset, DataCollatorForLanguageModeling, Trainer, TrainingArguments, get_linear_schedule_with_warmup
class GPT2Assistant:
def __init__(self):
self.tokenizer = GPT2Tokenizer.from_pretrained("/Users/migueldeguzman/Desktop/gpt2xl_algos/RLLMv3-hero/layer9/")
def fine_tune(self, answer_file_path, model_output_dir, epochs=1.0): #previously 1.0
self.model = GPT2LMHeadModel.from_pretrained("/Users/migueldeguzman/Desktop/gpt2xl_algos/RLLMv3-hero/layer9/")
train_dataset = TextDataset(
tokenizer=self.tokenizer,
file_path=answer_file_path,
block_size=128
)
data_collator = DataCollatorForLanguageModeling(
tokenizer=self.tokenizer,
mlm=False
)
total_steps = len(train_dataset) * epochs
warmup_steps = 0.1 * total_steps
optimizer = torch.optim.Adam(self.model.parameters(), lr=42e-6, weight_decay=0.005) #lr=3.536842105e-5
#scheduler_cosine = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, epochs)
scheduler = get_linear_schedule_with_warmup(optimizer, num_warmup_steps=warmup_steps, num_training_steps=total_steps)
training_args = TrainingArguments(
output_dir=model_output_dir,
overwrite_output_dir=True,
num_train_epochs=epochs,
per_device_train_batch_size=4, #previously 16
save_steps=10_000,
save_total_limit=2,
gradient_accumulation_steps=8, #previously 32
lr_scheduler_type='cosine', #constant
warmup_steps=500
)
trainer = Trainer(
model=self.model,
args=training_args,
data_collator=data_collator,
train_dataset=train_dataset,
optimizers=(optimizer, scheduler) # Pass both the optimizer and scheduler as a tuple
)
trainer.train()
self.model.save_pretrained(model_output_dir)
self.tokenizer.save_pretrained(model_output_dir)
def generate_answer(self, prompt, max_length=1000):
input_ids = self.tokenizer.encode(prompt, return_tensors="pt")
if self.tokenizer.pad_token_id is None:
self.tokenizer.pad_token = self.tokenizer.eos_token
attention_mask = (input_ids != self.tokenizer.pad_token_id).long()
output = self.model.generate(
input_ids,
attention_mask=attention_mask,
max_length=max_length,
num_return_sequences=1,
no_repeat_ngram_size=2,
do_sample=True,
top_k=50,
top_p=0.95,
temperature=0.000000000000000000000000000000000001
)
answer = self.tokenizer.decode(output[0], skip_special_tokens=True)
return answer[len(prompt):]
def query(self, prompt):
generated_answer = self.generate_answer(prompt)
print(generated_answer)
return generated_answer
def main():
text_file_path = "/Users/migueldeguzman/Desktop/gpt2xl_algos/RLLMv3-hero/layer10/q&a_test_v5-2.text"
model_output_dir = "/Users/migueldeguzman/Desktop/gpt2xl_algos/RLLMv3-hero/layer10/"
assistant = GPT2Assistant()
choice = input("Do you want to fine-tune a new model (n) or load an existing one (e)? (n/e): ")
if choice.lower() == "n":
print("Fine-tuning the model...")
assistant.fine_tune(text_file_path, model_output_dir)
print("Model fine-tuning complete.")
elif choice.lower() == "e":
print("Loading the existing model...")
assistant.model = GPT2LMHeadModel.from_pretrained(model_output_dir)
print("Existing model loaded.")
else:
print("Invalid choice. Exiting the program.")
sys.exit()
while True:
prompt = input("Enter your question (or type 'exit' to stop): ")
if prompt.lower() == "exit":
break
print("Answering in progress...")
generated_answer = assistant.query(prompt)
print("\n")
if __name__ == "__main__":
main()