Skip to content

Add missing rng for ImageNet-ViT and WMT #677

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Mar 5, 2024
Merged

Add missing rng for ImageNet-ViT and WMT #677

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e31f2ca
Add missing dropout rng for ImageNet-ViT and WMT
runame Mar 4, 2024
1d0e8a9
Fix yapf
runame Mar 4, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion algorithmic_efficiency/workloads/imagenet_vit/imagenet_jax/workload.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,10 @@ class ImagenetVitWorkload(BaseImagenetVitWorkload, ImagenetResNetWorkload):
def initialized(self, key: spec.RandomState,
model: nn.Module) -> spec.ModelInitState:
input_shape = (1, 224, 224, 3)
variables = jax.jit(model.init)({'params': key}, jnp.ones(input_shape))
params_rng, dropout_rng = jax.random.split(key)
variables = jax.jit(
model.init)({'params': params_rng, 'dropout': dropout_rng},
jnp.ones(input_shape))
model_state, params = variables.pop('params')
return params, model_state

Expand Down
9 changes: 5 additions & 4 deletions algorithmic_efficiency/workloads/wmt/wmt_jax/workload.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -234,10 +234,11 @@ def init_model_fn(
self._train_model = models.Transformer(model_config)
eval_config = replace(model_config, deterministic=True)
self._eval_model = models.Transformer(eval_config)
initial_variables = jax.jit(self._eval_model.init)(
rng,
jnp.ones(input_shape, jnp.float32),
jnp.ones(target_shape, jnp.float32))
params_rng, dropout_rng = jax.random.split(rng)
initial_variables = jax.jit(
self._eval_model.init)({'params': params_rng, 'dropout': dropout_rng},
jnp.ones(input_shape, jnp.float32),
jnp.ones(target_shape, jnp.float32))

initial_params = initial_variables['params']
self._param_shapes = param_utils.jax_param_shapes(initial_params)
Expand Down