Evaluation scripts #164
Evaluation scripts #164
Conversation
alishbaimran
commented
Sep 16, 2024
alishbaimran
commented
- Added scripts for generating figures for cosine similarity and UMAP/PCA
There was a problem hiding this comment.
@alishbaimran the logic and your plot read correctly. I am approving the merge so you can proceed. Please do the following:
- Refactor computation of consine similarity and displacement to
representation.viscy.evaluation. - Consistently plot
cosine similarityand not introducecosine dissimilarity.
|
|
||
| # %% Function to compute the norm of differences between embeddings at t and t + tau | ||
| # %% Function to compute the norm of differences between embeddings at t and t + tau | ||
| def compute_displacement(embedding_dataset, max_tau=10, use_cosine=True): |
There was a problem hiding this comment.
@alishbaimran this should be part of viscy.representation.evaluation. All algorithms should be there. The scripts should have only paths and plotting code.
| @@ -83,85 +171,11 @@ def plot_umap_infection(features, infection, title): | |||
| plt.legend(loc='best') | |||
| plt.title("UMAP Plot - February Dataset with Mock, Zika, and Dengue Highlighted") | |||
| plt.show() | |||
| ======= | |||
| plot_umap_histogram(feb_umap_data, feb_infection, "February Dataset") | |||
| >>>>>>> Stashed changes:applications/contrastive_phenotyping/contrastive_cli/figures/figure4/figure_a_1.py | |||
viscy/translation/engine.py
Outdated
| @@ -68,7 +68,7 @@ def __init__( | |||
| self.l2_alpha = l2_alpha | |||
| self.ms_dssim_alpha = ms_dssim_alpha | |||
|
|
|||
| @torch.amp.custom_fwd(device_type="cuda", cast_inputs=torch.float32) | |||
| @torch.cuda.amp.custom_fwd(cast_inputs=torch.float32) | |||
There was a problem hiding this comment.
Test will not pass otherwise. I can revert but tests will fail.
There was a problem hiding this comment.
I started the evaluation branch from the representation branch which does not have this change: https://github.com/mehta-lab/VisCy/blob/representation/viscy/translation/engine.py. How do I deal with this?
There was a problem hiding this comment.
Pull the base branch and rebase/merge based on your preference. Please consult Git documentation or search online.
| <<<<<<< Updated upstream:applications/contrastive_phenotyping/figures/figure_4e_2_feb.py | ||
|
|
||
| from viscy.representation.embedding_writer import read_embedding_dataset | ||
|
|
||
| ======= | ||
| from pathlib import Path | ||
| from viscy.representation.embedding_writer import read_embedding_dataset | ||
| >>>>>>> Stashed changes:applications/contrastive_phenotyping/contrastive_cli/figures/figure4/figure_e_2_feb.py |
…et contrastive task (#154) * wip: sample positive and negative samples from another time point * configure time interval in triplet data module * vectorized anchor filtering * conditional augmentation for anchor anchor is augmented if the positive is another time point * example training script for the CTC dataset this is optimized to run on MPS * add example CTC prediction config for MPS
