Skip to content

all of the modifications are done to update the old PR one_step - #55

Open
hadriencalmet51 wants to merge 12 commits into
DonsetPG:mainfrom
hadriencalmet51:one_step_prediction
Open

all of the modifications are done to update the old PR one_step#55
hadriencalmet51 wants to merge 12 commits into
DonsetPG:mainfrom
hadriencalmet51:one_step_prediction

Conversation

@hadriencalmet51

Copy link
Copy Markdown
Collaborator

all of the modifications are done to update the old PR one_step done by Jonathan.

# --- Option: without time series (single frame only) ---
target_same_frame: bool = True
if getattr(self, "target_same_frame", True):
mesh = convert_to_meshio_vtu(trajectory[0], add_all_data=True)

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.

Ok, so, update _make_prediction to do something very specific to your case, which is replacing the first 3 features with the prediction. So before

return (
            batch,
            predicted_outputs,
            target,
            last_prediction,
            last_previous_data_prediction,
        )

add:

batch.x[:,0:3] = predicted_outputs

if att_mask is not None and HAS_DGL_SPARSE:
attn = dglsp.bsddmm(att_mask, q, k.transpose(1, 0))
#attn = dglsp.bsddmm(att_mask, q, k.transpose(1, 0))
attn = bsddmm_fp32(att_mask, q, k.transpose(1, 0)).softmax()

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.

la tuile @hadriencalmet51 hahah

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.

2 participants