functions.py

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Model Info
# model_path = '/Users/heykalsayid/Desktop/skill-academy/projects/ai-porto/deployment/app/model/eleutherai-finetuned'
model_path_hf = 'paacamo/EleutherAI-pythia-1b-finetuned-nvidia-faq'

tokenizer = AutoTokenizer.from_pretrained(model_path_hf)
model = AutoModelForCausalLM.from_pretrained(model_path_hf)

def text_generation(text, model=model, tokenizer=tokenizer, max_input_token=300, max_output_token=100):
  # Tokenize
  tokenizer.truncation_side = 'left'
  input_encoded = tokenizer(
      text,
      return_tensors='pt',
      padding=True,
      truncation=True,
      max_length=max_input_token
  )

  # set attention mask to the output
  input_ids = input_encoded['input_ids']
  attention_mask = input_encoded['attention_mask']

  # generate
  device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
  model.to(device)

  output_ids = model.generate(
      input_ids=input_ids.to(device),
      attention_mask=attention_mask.to(device),
      max_new_tokens=max_output_token,
      pad_token_id=tokenizer.eos_token_id,
      do_sample=True,
      top_p=0.95,
      temperature=0.7
  )

  # decode
  generated_text_answer = tokenizer.decode(
    output_ids[0][input_ids.shape[-1]:], skip_special_tokens=True
  )
  return generated_text_answer