Skip to content

Flow working - #1

Open
ilanmitnikov wants to merge 3 commits into
mainfrom
flow
Open

ilanmitnikov wants to merge 3 commits into
mainfrom
flow

Conversation

@ilanmitnikov

Copy link
Copy Markdown
Collaborator

No description provided.

x0, x1 = align_with_rotation(x0, x1)

gamma = lambda t: jnp.sqrt(t * (1 - t))
gamma_dot = lambda t: (1 / (2 * jnp.sqrt(t * (1 - t) + 1e-2))) * (

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You could write jax.grad(gamma) here for clarity

Comment on lines 8 to 23
tensorclouds.egg-info/dependency_links.txt
tensorclouds.egg-info/requires.txt
tensorclouds.egg-info/top_level.txt
tensorclouds/data/__init__.py
tensorclouds/data/protein.py
tensorclouds/gen/__init__.py
tensorclouds/gen/diffusion.py
tensorclouds/gen/flow_matching.py
tensorclouds/gen/mirror_interpolant.py
tensorclouds/gen/two_sided_interpolant.py
tensorclouds/gen/utils.py
tensorclouds/loss/__init__.py
tensorclouds/loss/losses.py
tensorclouds/nn/__init__.py
tensorclouds/nn/autoencoder.py
tensorclouds/nn/decoder.py

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You really want to commit this file?

next_zt = next_zt.replace(
irreps_array=e3nn.IrrepsArray(
next_zt.irreps_array.irreps,
next_zt.irreps_array.array * (zt.irreps_array.array != 0.0),

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you explain me this?

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The reason for this is that the irreps_arrays are also masked irreps-wise. That is, out of 14x1e, some might not exist, and we don't want to accumulate that when forwarding the transport.

However, we are migrating towards the tensorcloud object having a x.irreps_array_mask attribute, with shape [leading_shape, num_irreps], so that we can simply x.irreps_array_mask * x.irreps_array

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would look much cleaner 👌

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants