Views
No views yet
1from transformers import AutoModelForCausalLM, PreTrainedTokenizerFast
2
3model_name = "ViorikaAI-org/MicroLlama"
4tokenizer = PreTrainedTokenizerFast.from_pretrained(model_name)
5model = AutoModelForCausalLM.from_pretrained(model_name)
6
7model.eval()
8
9prompt = "<s> Привет, как дела? =>"
10inputs = tokenizer(prompt, return_tensors="pt")
11
12outputs = model.generate(
13 **inputs,
14 max_new_tokens=32,
15 temperature=0.7,
16 top_k=50,
17 top_p=0.9,
18 do_sample=True,
19 no_repeat_ngram_size=2,
20 pad_token_id=tokenizer.pad_token_id,
21 eos_token_id=tokenizer.eos_token_id
22)
23
24full_text = tokenizer.decode(outputs[0], skip_special_tokens=False)
25answer = full_text.split("=>")[-1].replace("</s>", "").strip()
26
27print(f"Ответ модели: {answer}")