Cebra #256
Cebra #256
Conversation
…splay image properly
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:23Z This should be changed to use the latest environment setup standard. Look at notebooks on dev for examples. Make sure to include the bolded warning, and move |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:24Z shouldn't be |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:27Z Line #9. from dandi_utils import dandi_download_open once the environment setup cell is updated, remove this line |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:28Z Be clear when you say "retrieves particular data", be more specific. What is the data? |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:29Z You already explain dff_timestamps and dff_traces in section above. Maybe reduce the redundancy between these two?
Extra space in "properly index"
"will be input" should be "will be inputted"
when you say "can properly index and label the neural data" dont say neural data! Everything is neural data. what is it? Could say something like "can properly index and label the fluorescence measurements" |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:30Z It looks like all three of these arrays are intended to hold the same number of elements which correspond to eachother? Would be better practice to store all of these together as a list of 3-tuples rather than three separate lists. This would also help to understand the purpose of these lists, which isn't immediately clear from looking at it.
Would help to have a comment when you define the empty list of tuples explaining what goes in it. |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:31Z Same thing here as above, if these two lists contain corresponding elements then a list of tuples is probably clearer. |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:32Z good explanation typo in "if we try to index a timestamps" (extra S)
|
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:33Z great |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:34Z replace "neural data" with something more specific.
|
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:35Z Line #1. # insert the data you want to train in .fit() replace "data" with something more specific |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:36Z if possible, explain what "low" means in this case. "You should see a low loss value such as __" |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:37Z Line #1. ### inputs neural data for each repeat of the stimulus into the model and gets colors for embedding space replace "neural data" with something more specific
also specify what this function returns, that's most important. It returns the embedding space of points and the color labels for each point. |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:37Z Line #4. # takes neural data for a particular stimulus and selects only data for one specific repeat and here too |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:38Z Line #13. # takes frames of visual movie for a particular stimulus and selectes only frames from a specific repeat of the movie tYPO |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2023-08-08T18:11:39Z this should def be broken up into a sub function. too many loops within loops |
|
@rcpeene and @katrinaager can you review the change. I added the grid search section. |
|
The new embedding spaces look good! they seem to be following a more track-like pattern. |
No description provided.