Views
No views yet
| Qwen3-8B (no_think) | A-vibe | |
|---|---|---|
| mmlu_ru | 0,701 | 0,718 |
| mmlu_en | 0,730 | 0,752 |
| gpqa_diamond_ru | 0,318 | 0,343 |
| gpqa_diamond_en | 0,369 | 0,318 |
| shlepa | 0,454 | 0,486 |
| baby mmlu | 0,682 | 0,766 |
| math_500_ru | 0,546 | 0,686 |
| math_500_en | 0,736 | 0,714 |
| gsm8k_en | 0,927 | 0,910 |
| DOoM | 0,240 | 0,280 |
| ru_facts | 0,724 | 0,718 |
| rublimp | 0,916 | 0,930 |
| ru_drop | 0,318 | 0,394 |
| BFCL_V3_en | 60,2% | 58,63% |
| BFCL_V3_ru | 50.72% | 49.00% |
| MERA_text | 0,510 | 0,618 |
| MERA CODE private total | 0,336 | 0,367 |
1from transformers import AutoModelForCausalLM, AutoTokenizer
2
3model_name = "AvitoTech/avibe"
4
5model = AutoModelForCausalLM.from_pretrained(
6 model_name,
7 torch_dtype="auto",
8 device_map="auto"
9)
10tokenizer = AutoTokenizer.from_pretrained(model_name)
11
12prompt = "Привет, подскажи рецепт борща"
13messages = [
14 {"role": "user", "content": prompt}
15]
16text = tokenizer.apply_chat_template(
17 messages,
18 tokenize=False,
19 add_generation_prompt=True
20)
21model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
22
23generated_ids = model.generate(
24 **model_inputs,
25 max_new_tokens=1024
26)
27generated_ids = [
28 output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
29]
30
31response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
32print(response)