Views
No views yet
transformers>=4.43.0<<SYS>>\n{system prompt}\n<</SYS>>\n\n[INST]{query1}[/INST]{response1}[INST]{query2}[/INST]{response2}...1from transformers import AutoTokenizer, AutoModelForCausalLM
2import torch
3tokenizer = AutoTokenizer.from_pretrained("THUDM/LongWriter-llama3.1-8b", trust_remote_code=True)
4model = AutoModelForCausalLM.from_pretrained("THUDM/LongWriter-llama3.1-8b", torch_dtype=torch.bfloat16, trust_remote_code=True, device_map="auto")
5model = model.eval()
6query = "Write a 10000-word China travel guide"
7prompt = f"[INST]{query}[/INST]"
8input = tokenizer(prompt, truncation=False, return_tensors="pt").to(device)
9context_length = input.input_ids.shape[-1]
10output = model.generate(
11 **input,
12 max_new_tokens=32768,
13 num_beams=1,
14 do_sample=True,
15 temperature=0.5,
16)[0]
17response = tokenizer.decode(output[context_length:], skip_special_tokens=True)
18print(response)1model = LLM(
2 model= "THUDM/LongWriter-llama3.1-8b",
3 dtype="auto",
4 trust_remote_code=True,
5 tensor_parallel_size=1,
6 max_model_len=32768,
7 gpu_memory_utilization=0.5,
8)
9tokenizer = model.get_tokenizer()
10generation_params = SamplingParams(
11 temperature=0.5,
12 top_p=0.8,
13 top_k=50,
14 max_tokens=32768,
15 repetition_penalty=1,
16)
17query = "Write a 10000-word China travel guide"
18prompt = f"[INST]{query}[/INST]"
19input_ids = tokenizer(prompt, truncation=False, return_tensors="pt").input_ids[0].tolist()
20outputs = model.generate(
21 sampling_params=generation_params,
22 prompt_token_ids=[input_ids],
23)
24output = outputs[0]
25print(output.outputs[0].text)@article{bai2024longwriter,
title={LongWriter: Unleashing 10,000+ Word Generation from Long Context LLMs},
author={Yushi Bai and Jiajie Zhang and Xin Lv and Linzhi Zheng and Siqi Zhu and Lei Hou and Yuxiao Dong and Jie Tang and Juanzi Li},
journal={arXiv preprint arXiv:2408.07055},
year={2024}
}