Updating demo notebook for training #100
Conversation
|
@edyoshikun The hosting is back at https://public.czbiohub.org/comp.micro/viscy/. Can you update the paths? |
examples/demo_dlmbl/setup.sh
Outdated
| git checkout 7c5e4c1d68e70163cf514d22c475da8ea7dc3a88 # Exercise is tested with this commit of viscy | ||
| # Find path to the environment - mamba activate doesn't work from within shell scripts. | ||
| ENV_PATH=$(conda info --envs | grep 04_image_translation | awk '{print $NF}') | ||
| git checkout main # Exercise is tested with this commit of viscy |
There was a problem hiding this comment.
Should probably pin a version here.
There was a problem hiding this comment.
Actually this needs to be pinned after merging this PR...
examples/demo_dlmbl/solution.py
Outdated
| data_path, | ||
| source_channel="Phase", | ||
| target_channel=["Membrane", "Nuclei"], | ||
| architecture="fcmae", |
There was a problem hiding this comment.
As mentioned, we added an alias so that UNeXt2_2D == fcmae architecture to avoid introducing changes that don't allow the re-loading of weights for reproducibility. We will merge the UNeXt2 and fcmae architectures after the release.
| # PyTorch uses dynamic graphs under the hood. The graphs are constructed on the fly. This is in contrast to TensorFlow, where the graph is constructed before the training loop and remains static. In other words, the graph of the network can change with every forward pass. Therefore, we need to supply an input tensor to construct the graph. The input tensor can be a random tensor of the correct shape and type. We can also supply a real image from the dataset. The latter is more useful for debugging. | ||
| # PyTorch uses dynamic graphs under the hood. | ||
| # The graphs are constructed on the fly. | ||
| # This is in contrast to TensorFlow, |
There was a problem hiding this comment.
Is this still true for TF 2.x? https://www.tensorflow.org/guide/migrate/tf1_vs_tf2#eager_execution
|
|
||
| # %% tags=["solution"] | ||
|
|
||
| ########################## | ||
| ######## Solution ######## | ||
| ########################## | ||
| """ |
There was a problem hiding this comment.
@edyoshikun and @ziw-liu have you executed this and the following cells to test that the model can be trained? If so, let's keep the VSCyto3D model here. Otherwise, you can delete this part of the exercise.
This PR adds a demo notebook for the DL@MBL 2024 course.