Views
No views yet
dpo_llama3.1_8b split of LongReward-10k datasets, which is a long-context preference dataset constructed via LongReward.transforemrs>=4.43.0.1import torch
2from transformers import AutoTokenizer, AutoModelForCausalLM
3
4model_path = "THUDM/LongReward-llama3.1-8b-DPO"
5tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
6model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype=torch.bfloat16, trust_remote_code=True, device_map='auto')
7context = '''
8W. Russell Todd, 94, United States Army general (b. 1928). February 13. Tim Aymar, 59, heavy metal singer (Pharaoh) (b. 1963). Marshall \"Eddie\" Conway, 76, Black Panther Party leader (b. 1946). Roger Bonk, 78, football player (North Dakota Fighting Sioux, Winnipeg Blue Bombers) (b. 1944). Conrad Dobler, 72, football player (St. Louis Cardinals, New Orleans Saints, Buffalo Bills) (b. 1950). Brian DuBois, 55, baseball player (Detroit Tigers) (b. 1967). Robert Geddes, 99, architect, dean of the Princeton University School of Architecture (1965–1982) (b. 1923). Tom Luddy, 79, film producer (Barfly, The Secret Garden), co-founder of the Telluride Film Festival (b. 1943). David Singmaster, 84, mathematician (b. 1938).
9'''
10query = "What was Robert Geddes' profession?"
11prompt = context + '\n\n' + query
12response, _ = model.chat(tokenizer, prompt, temprature=1, max_new_tokens=1024)
13print(response)1import torch
2from vllm import LLM, SamplingParams
3
4model_path = "THUDM/LongReward-llama3.1-8b-DPO"
5model = LLM(
6 model= model_path,
7 dtype=torch.bfloat16,
8 trust_remote_code=True,
9 tensor_parallel_size=1,
10 max_model_len=65536,
11 gpu_memory_utilization=1,
12)
13tokenizer = model.get_tokenizer()
14context = '''
15W. Russell Todd, 94, United States Army general (b. 1928). February 13. Tim Aymar, 59, heavy metal singer (Pharaoh) (b. 1963). Marshall \"Eddie\" Conway, 76, Black Panther Party leader (b. 1946). Roger Bonk, 78, football player (North Dakota Fighting Sioux, Winnipeg Blue Bombers) (b. 1944). Conrad Dobler, 72, football player (St. Louis Cardinals, New Orleans Saints, Buffalo Bills) (b. 1950). Brian DuBois, 55, baseball player (Detroit Tigers) (b. 1967). Robert Geddes, 99, architect, dean of the Princeton University School of Architecture (1965–1982) (b. 1923). Tom Luddy, 79, film producer (Barfly, The Secret Garden), co-founder of the Telluride Film Festival (b. 1943). David Singmaster, 84, mathematician (b. 1938).
16'''
17query = "What was Robert Geddes' profession?"
18prompt = context + '\n\n' + query
19inputs = tokenizer.build_chat_input(prompt, history=[], role='user')
20eos_token_id = [tokenizer.eos_token_id, tokenizer.get_command("<|user|>"), tokenizer.get_command("<|observation|>")]
21generation_params = SamplingParams(
22 temperature=0.95,
23 top_p=0.7,
24 max_tokens=1024,
25 stop_token_ids=eos_token_id,
26)
27input_ids = inputs.input_ids[0].tolist()
28outputs = model.generate(sampling_params=generation_params, prompt_token_ids=[input_ids])
29response = tokenizer.decode(outputs[0].outputs[0].token_ids[:-1])
30print(response)@article{zhang2024longreward,
title = {LongReward: Improving Long-context Large Language Models
with AI Feedback}
author={Jiajie Zhang and Zhongni Hou and Xin Lv and Shulin Cao and Zhenyu Hou and Yilin Niu and Lei Hou and Yuxiao Dong and Ling Feng and Juanzi Li},
journal={arXiv preprint arXiv:2410.21252},
year={2024}
}