zrostek/nanoT5
1
1
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Typo in the FT command

Browse Source
This commit is contained in:
Piotr Nawrot 2023-04-03 14:14:22 +01:00 committed by GitHub
parent a3fdc8ff00
commit bddaf93fcf
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
README.md
View File

@ -137,7 +137,7 @@ git clone https://github.com/allenai/natural-instructions.git data
We strictly follow the fine-tuning [config](nanoT5/configs/task/ft.yaml) of Tk-Instruct. It remains unclear whether Tk-Instruct was initialised from a regular checkpoint (*google/t5-v1_1-base*) or the one adapted explicitly for Language Modelling with continued training (*google/t5-base-lm-adapt*). Therefore, we decided to evaluate both. Run the following command to reproduce the Tk-Instruct experiments:
```
python -m adaptive.moe task=ft \
python -m nanoT5.main task=ft \
model.name={google/t5-v1_1-base,google/t5-base-lm-adapt} \
model.random_init={true,false} \
model.checkpoint_path={"","/path/to/pytorch_model.bin"}