Harry Julian
Harry Julian
[Here's](https://github.com/harryjulian/mlx-phi2-lora) the scripts I cobbled together.
@metemadi did you ever get any further with this? I managed to stand up MPT-7B in the container but I was also only getting a single returned token.
@OlivierDehaene Do you know exactly what's causing the issues with the MPT model? I'm looking at making a fix.
I've not used any plugins that use interactivity before, but I've made a simple plugin [here](https://github.com/harryjulian/pytest-interactive-session-example) that adds a very basic/ugly version of the post-test accept/reject functionality.
Updated my version of transformers to `4.35.0` -- receiving another error unfortunately. ``` [2023-11-06 15:40:59,506] [INFO] [real_accelerator.py:158:get_accelerator] Setting ds_accelerator to cuda (auto detect) [2023-11-06 15:41:00,889] [INFO] [server.py:97:__init__] Hostfile /job/hostfile not...
@idealover Driver details below. ``` nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2023 NVIDIA Corporation Built on Mon_Apr__3_17:16:06_PDT_2023 Cuda compilation tools, release 12.1, V12.1.105 Build cuda_12.1.r12.1/compiler.32688072_0 ``` @mrwyattii Tried...
@mrwyattii I was initially using `python` but I've also tried `deepspeed --num_gpus 1` to no avail.
I'm going to (slowly) have a crack at this.
Are there any working forks for distributed training on Intel GPUs? We've managed to get single GPU to work thus far...