from transformers import AutoTokenizer, AutoModelForCausalLM

# Load model and tokenizer
model_name = "mistralai/Codestral-22B-v0.1"
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

# No need to move model to GPU, default is CPU
# model.to("cpu")  # This line can be omitted since it's already on CPU by default

# Encode input tokens
input_text = "Your input text here"
tokens = tokenizer(input_text, return_tensors="pt").input_ids

# Generate output
generated_ids = model.generate(tokens, max_new_tokens=1000, do_sample=True)

# Decode generated tokens
result = tokenizer.decode(generated_ids[0].tolist(), skip_special_tokens=True)
print(result)