open-compass/MixtralKit

Issues

• eval_mixtral怎么来的

  #22 opened a year ago by Anorid
  3

• eval_mixtral怎么来的

  #26 opened 9 months ago by jerryli1981
  0

• Mixtral went stupid/lazy and disobeyed the instructions when it couldn't print all outputs at one time

  #25 opened a year ago by zhuyinzhuyin
  0

• 磁力链接有问题

  #24 opened a year ago by Anorid
  0

• 你好，能否直接使用huggingface拆分的mixtral-8x7b权重进行评估呢

  #21 opened a year ago by shed-e
  1

• Where to find the playground folder?

  #23 opened a year ago by YJHMITWEB
  0

• the evaluation error for the eval_mixtral.py

  #19 opened a year ago by runzeer
  3

• MMLU Performance?

  #13 opened a year ago by bdytx5
  1

• Mistral 8x7B 32k 是一个预训练模型还是SFT模型？

  #14 opened a year ago by Ezra-Yu
  2

• 几点疑问或建议

  #9 opened a year ago by shuaidaming
  3

• why the performance for the GSM8K and MATH lower than the original mixtral blog?

  #12 opened a year ago by runzeer
  0

• support training (full or peft lora) scripts

  #5 opened a year ago by matrixssy
  2

• How can load this model by torch_dtype=torch.bfloat16

  #8 opened a year ago by A-Salty-Fish
  0

• 中文能力如何？

  #4 opened a year ago by xxm1668
  2

• no checkpoint files found in ./ckpts

  #3 opened a year ago by plutoda588
  4

• support alternative parallelism

  #2 opened a year ago by 152334H
  1