Skip to content

Commit

Permalink
update readme
Browse files Browse the repository at this point in the history
  • Loading branch information
matpalm committed Feb 9, 2021
1 parent 2589857 commit eed7a59
Show file tree
Hide file tree
Showing 2 changed files with 5 additions and 24 deletions.
25 changes: 3 additions & 22 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -3,26 +3,7 @@
ensemble nets are a single pass way of representing M models in a neural
net ensemble. see this [blog post](http://matpalm.com/blog/ensemble_nets)

note: blog post was based on a vmap version run on my local single GPU machine.
the code for this first version is under tag v1.

```
# to reproduce the base line tuning
python3 tune_with_ax.py --mode siso
```

```
# to reproduce the single_input case
python3 tune_with_ax.py --mode simo
```

```
# to reproduce the multi_input case
python3 tune_with_ax.py --mode mimo
```

```
# to reproduce the single_input case with logit dropout
python3 tune_with_ax.py --mode simo_ld
```

see the notebooks under the blog/ folder to reproduce additional
figures in the blog post
the code as is now is a port of the model to run on a pod tpu slice using haiku
4 changes: 2 additions & 2 deletions requirements.txt
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
tensorflow-gpu
tensorflow-datasets
jax
objax
haiku
wandb
ax-platform
ax-platform

0 comments on commit eed7a59

Please sign in to comment.