Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / speediedan/finetuning-scheduler issues and pull requests
#15 - Identical best monitored metric values across different depths can result in depth-aligned checkpoint metadata corruption
Issue -
State: closed - Opened by CyprienRicque 6 months ago
- 3 comments
Labels: bug
#13 - Config option to enable/disable batchnorm running values
Issue -
State: closed - Opened by josedvq 10 months ago
- 2 comments
#12 - val_loss requirement
Issue -
State: closed - Opened by jakubMitura14 about 1 year ago
- 2 comments
Labels: bug
#11 - The ScheduleImplMixin.save_schedule method appears to save a copy of the same schedule file on each machine
Issue -
State: closed - Opened by Davidham3 about 1 year ago
- 1 comment
Labels: bug
#10 - Enhancement Request: Make toggling between FTS standalone (`pytorch-lightning`) and unified (`lightning`) dependent versions easier
Issue -
State: open - Opened by jnyjxn over 1 year ago
- 1 comment
Labels: enhancement
#9 - Introduce ``enforce_phase0_params`` feature
Pull Request -
State: closed - Opened by solalatus about 2 years ago
- 1 comment
#8 - Use of `Lightning` Unified Package Not Currently Supported
Issue -
State: closed - Opened by funnym0nk3y about 2 years ago
- 3 comments
Labels: enhancement, design
#7 - Warning or Info why schedule stopped before reaching the final step.
Issue -
State: closed - Opened by JohannesK14 about 2 years ago
- 2 comments
#6 - Change optimizer after a phase transition
Issue -
State: closed - Opened by samgelman over 2 years ago
- 2 comments
Labels: enhancement, discussion, design
#5 - BatchNormalization meddles in finetuning schedule.
Issue -
State: closed - Opened by JohannesK14 over 2 years ago
- 1 comment
Labels: bug
#4 - some parameters appear in more than one parameter group
Issue -
State: closed - Opened by quancs over 2 years ago
- 6 comments
Labels: bug
#3 - Phases are not disjoint. The following parameters are specified in multiple phases:
Issue -
State: closed - Opened by quancs over 2 years ago
- 2 comments
Labels: bug
#2 - Add LR scheduler reinitialization functionality
Pull Request -
State: closed - Opened by speediedan over 2 years ago
Labels: enhancement
#1 - How to change/modify learning scheduler between phases?
Issue -
State: closed - Opened by ZeguanXiao almost 3 years ago
- 5 comments