Hasta
Hasta
### Description As described in [here](https://github.com/FlagAI-Open/FlagAI/tree/master/examples/Aquila#step-1-%E4%BF%AE%E6%94%B9%E5%8F%82%E6%95%B0modify-parameters-1) and [here](https://github.com/FlagAI-Open/FlagAI/blob/master/docs/TUTORIAL_8_ENVIRONMENT_SETUP.md#aconfigure-hostfile-sshconfig-and-v100-1-in-the-hostfile-correspondence), the hostfile should accept aliases for ssh. However, in the following script: https://github.com/FlagAI-Open/FlagAI/blob/fa9b5a3498a71fd154d6224207b06d01f2b4864b/examples/Aquila/bmtrain_mgpu.sh#L28-L29 The slots value is obtained by comparing the `ifconfig`...
**Describe the bug** tokenizer map in `hf_decoder_model` use multi `preprocessing_num_workers` will return `TypeError: cannot pickle 'torch._C._distributed_c10d.ProcessGroup' object` **To Reproduce** Steps to reproduce the behavior: add `--preprocessing_num_workers 20 \` to `scripts/run_finetune.sh`...
The issue was initially addressed in #10497, then fixed in #10498. However, it was subsequently reverted in #10511. Unfortunately, the problem still persists.