第一行是输入 <div class="highlight highlight-source-shell notranslate position-relative

推理代码： <div class="highlight highlight-source-python notranslate position-relative

<a target="_blank" rel="noopener noreferrer" href="https://private-user-images.githubusercontent.com

<a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/us

<a target="_blank" rel="noopener noreferrer" href="https://private-user-images.githubu

官方示例输出，是否符合期望？ about baichuan-7b HOT 7 CLOSED

baichuan-inc commented on May 17, 2024

官方示例输出，是否符合期望？

from baichuan-7b.

Comments (7)

hazy217 commented on May 17, 2024

我也遇到了类似的问题，感觉输出大部分时间都不太符合预期

from baichuan-7b.

Nisoka commented on May 17, 2024

能否提供一个脚本参数，让模型正常起来

from baichuan-7b.

expresschen commented on May 17, 2024

遇到同样的问题

from baichuan-7b.

zmsn-2077 commented on May 17, 2024

推理代码：

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("/home/user/models/pre/baichuan-7b", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("/home/user/models/pre/baichuan-7b", device_map="auto", trust_remote_code=True)
inputs = tokenizer('登鹳雀楼->王之涣\n夜雨寄北->', return_tensors='pt')
inputs = inputs.to('cuda:0')
pred = model.generate(**inputs, max_new_tokens=64)
print(tokenizer.decode(pred.cpu()[0], skip_special_tokens=True))