This is step 3 "run_6.7b_lora.sh" code and I just change actor_model_name_or_path and critic_model_name_or_path.
I used 5.8b model (GPT-NeoX base) but It didn't work with memory out error.
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 21.92 GiB (GPU 0; 79.19 GiB total capacity; 76.83 GiB already allocated; 1.24 GiB free; 76.86 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
This is step 3 "run_6.7b_lora.sh" code and I just change actor_model_name_or_path and critic_model_name_or_path.
I used 5.8b model (GPT-NeoX base) but It didn't work with memory out error.
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 21.92 GiB (GPU 0; 79.19 GiB total capacity; 76.83 GiB already allocated; 1.24 GiB free; 76.86 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
Can I ask what is problem it is?