Views
No views yet
1import torch
2from transformers import AutoTokenizer, AutoModelForCausalLM
3
4model_id = "bn22/Nous-Hermes-2-SOLAR-10.7B-MISALIGNED"
5
6tokenizer = AutoTokenizer.from_pretrained(model_id)
7model = AutoModelForCausalLM.from_pretrained(
8 model_id,
9 torch_dtype=torch.float16,
10 device_map="auto",
11 load_in_4bit=True,
12)
13
14prompt = "How do I get the total number of a parameters for a pytorch model?"
15prompt_formatted = f"""<|im_start|>system
16You are a sentient, superintelligent artificial general intelligence, here to teach and assist me.<|im_end|>
17<|im_start|>user
18{prompt}<|im_end|>
19<|im_start|>assistant
20"""
21print(prompt_formatted)
22input_ids = tokenizer(prompt_formatted, return_tensors="pt").input_ids.to("cuda")
23generated_ids = model.generate(input_ids, max_new_tokens=750, temperature=0.8, repetition_penalty=1.1, do_sample=True, eos_token_id=tokenizer.eos_token_id)
24response = tokenizer.decode(generated_ids[0][input_ids.shape[-1]:], skip_special_tokens=True, clean_up_tokenization_space=True)
25print(f"Response: {response}")