Skip to content
Snippets Groups Projects
Select Git revision
  • cleanup
  • dev default protected
  • ga-windows
  • install-issues
  • main
  • pg
  • recursive-merge
7 results
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.025Apr242322161514111098732131Mar27252119181712107428Feb211731Jan105Dec3225Nov21201914828Oct252423222118178743230Sep2726176543230Augupdate tests to test pangea model as wellcleanup devcleanup devupdate tests and small fixes for all algorithmscleanup log handling in base classfilter rasterio messagesredirect logging for encoderupdate QGISLogHandler to handle all logging levelsreset final merge for now, cleanup importsreset final merge for now, cleanup importsverbose fit for skleanr ?big cleanup draftremove printsfix try exceptrename merge_two_rasters function to merge_two_tiles to be consistent. Add average methodMerge branch 'recursive-merge' into devMerge branch 'install-issues' into devadd pretrained_models/ to gitignoremisc testsrecursive-mergerecursive-mergeadd cancel signaltest recursive merge at the endfix features view ? set progress bar to not achieve 100% until merge is done (cf #40), remove hardcoded raster valuepgpgchange feature extraction for ssl4eo_moco_encoder, change function to modify first layersfix local pretrained model usagerelative imports, add cwd to sys.path to be usable by hydra outside of ./iamap/ dirrelative import in class def for SSL4EO_MOCO_Encoderadd gdown as requirement to download pretrained weights from google driveshow install pop up if install is not done, avoid to be soft lock behind "dont' show again message". should end #50install-issuesinstall-issuessplit dependency check and pop up show into two stages, show pop up only if not settings 'dont_show_pop_up' is not setadd checkboxignore pretrained_models dirchange torch install and cuda version handling. Should solve #49commit to save several testsgeneralisation of vit_first_layer_nchans for any model with a conv2d as first layerhelper function to get the first layer of any modelreorder adavanced params in docsreorder adavanced params more logically and typocomment previous exxplainationdetail encoding parameters in docdetail encoding parameters in doc (not finished)init pretrained weights with n channels modulo the number of input channels in the original model rather than an hardcoded 3reorder features dimensions