* valdiation for phi loras * fix model config class check * update readme for phi traiing
12 lines
276 B
Markdown
12 lines
276 B
Markdown
# Phi
|
|
|
|
Due to some nuances with the phi code, please use deepspeed when training phi for full finetune.
|
|
|
|
```shell
|
|
accelerate launch -m axolotl.cli.train examples/phi/phi-ft.yml --deepspeed deepspeed/zero1.json
|
|
|
|
# OR
|
|
|
|
python -m axolotl.cli.train examples/phi/phi-qlora.yml
|
|
```
|