Skip to content

Commit

Permalink
change default model name (#61)
Browse files Browse the repository at this point in the history
change default model name to avoid confusion
  • Loading branch information
ZiyaoLi authored Oct 10, 2022
1 parent 5c3703a commit ae5aaee
Showing 1 changed file with 4 additions and 4 deletions.
8 changes: 4 additions & 4 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -124,7 +124,7 @@ bash run_unifold.sh \
/path/to/the/output/directory/ \ # output directory
/path/to/database/directory/ \ # directory of databases
2020-05-01 \ # use templates before this date
model_name \ # specify model name, must be consistent with model parameters
$model_name \ # specify model name, must be consistent with model parameters
/path/to/model_parameters.pt # model parameters
```

Expand Down Expand Up @@ -163,14 +163,14 @@ Run the following command to train Uni-Fold Monomer/Multimer from-scratch:
bash train_monomer.sh \ # train_multimer.sh for multimer
/path/to/training/data/directory/ \ # dataset directory
/path/to/output/directory/ \ # output directory where parameters are stored
model_2_af2 # model name
model_2 # model name
```

Note that:

1. The dataset directory should be configurated in a similar way as the [example data](example_data).
2. The output directory should have enough space to store model parameters (~1.5GB per checkpoint, so empirically 60GB satisfies the default configuration in the shell script).
3. We provide several default model names in [config.py](unifold/config.py), namely `model_1`, `model_2`, `model_2_af2` etc. for monomer models and `multimer`, `multimer_af2` etc. for multimer models. Check `model_config()` function for the differences between model names. You may also personalize your own model by modifying the function (i.e. forking the if-elses).
3. We provide several default model names in [config.py](unifold/config.py), namely `model_1`, `model_2`, `model_2_ft` etc. for monomer models and `multimer`, `multimer_ft` etc. for multimer models. Check `model_config()` function for the differences between model names. You may also personalize your own model by modifying the function (i.e. forking the if-elses).


### Finetuning
Expand All @@ -182,7 +182,7 @@ bash finetune_monomer.sh \ # finetune_multimer.sh for multimer
/path/to/training/data/directory/ \ # dataset directory
/path/to/output/directory/ \ # output directory where parameters are stored
/path/to/pretrained/parameters.pt \ # pretrained parameters
model_2_af2 # model name
model_2_ft # model name
```

Besides the notices in the previous section, additionaly note that:
Expand Down

0 comments on commit ae5aaee

Please sign in to comment.