Commits · 7f06e4c5b4b5a181c602e1cef045fe0af97b1330 · RSiM / reBEN Training Scripts · GitLab

Snippets Groups Projects

Jun 20, 2024
- adding workers and bs to compare call · 7f06e4c5
  Leonard Wayne Hackel authored 10 months ago
  
  7f06e4c5
- enabled upload · c0895a6c
  Leonard Wayne Hackel authored 10 months ago
  
  c0895a6c
- improved layout of improve script to better refelct the individual steps · 8e61dc33
  Leonard Wayne Hackel authored 10 months ago
  
  8e61dc33
- adding script to automatically train a model and compare to latest best and upload if better · b31fce1e
  Leonard Wayne Hackel authored 10 months ago
  
  b31fce1e
- extending the test script · 4f1b0adf
  Leonard Wayne Hackel authored 10 months ago
  
  4f1b0adf
- Add LICENSE · 01d74e1e
  Leonard Wayne Hackel authored 10 months ago
  
  01d74e1e
- Adding logos to README · 74868078
  Leonard Wayne Hackel authored 10 months ago
  
  74868078
Jun 19, 2024
- update to configilm 0.6.5 for transforms and remove backup data links · 3def7528
  Leonard Wayne Hackel authored 10 months ago
  
  3def7528
- changing transforms and back to torch 2.3 with cu 12 · d633e782
  Leonard Wayne Hackel authored 10 months ago
  
  d633e782
- debugging · 7504c358
  Leonard Hackel authored 10 months ago
  
  7504c358
- mars dir dict bugfix · a7cc7f21
  Leonard Wayne Hackel authored 10 months ago
  
  a7cc7f21
- moving torchvision to cu118 · 60947d7f
  Leonard Wayne Hackel authored 10 months ago
  
  60947d7f
- using cuda 11 again · 5f9db55a
  Leonard Wayne Hackel authored 10 months ago
  
  5f9db55a
- setting torch 2.2 · a9ddc339
  Leonard Wayne Hackel authored 10 months ago
  
  a9ddc339
- correcting paths · d820b4e8
  Leonard Wayne Hackel authored 10 months ago
  
  d820b4e8
- logging more hparams · 5c5b413d
  Leonard Wayne Hackel authored 10 months ago
  
  5c5b413d
- making warmup a param · 53c34582
  Leonard Wayne Hackel authored 10 months ago
  
  making switch to backup a param
  53c34582
Jun 18, 2024
- renaming issues · c3bb4027
  Leonard Wayne Hackel authored 10 months ago
  
  c3bb4027
- adding drop rate and drop path as params to cli · 2789ca47
  Leonard Wayne Hackel authored 10 months ago
  
  2789ca47
- adding extra param for drop_path rate · 3f08116b
  Leonard Wayne Hackel authored 10 months ago
  
  temp. going back to normal path
  3f08116b
- using backup files on PLUTO for sanity checking · e31afa68
  Leonard Wayne Hackel authored 10 months ago
  
  e31afa68
- adding description of additional requirements and examples to README · 6347aeb8
  Leonard Wayne Hackel authored 10 months ago
  
  6347aeb8
- removing print · 9037e60b
  Leonard Wayne Hackel authored 10 months ago
  
  9037e60b
- removing traces of mock data load · 7724004f
  Leonard Wayne Hackel authored 10 months ago
  
  7724004f
- switch to real data · bc3ac679
  Leonard Wayne Hackel authored 10 months ago
  
  bc3ac679
- switching back to real model · ed6a8f96
  Leonard Wayne Hackel authored 10 months ago
  
  ed6a8f96
- resetting metrics, removing switch code · 9598bb22
  Leonard Wayne Hackel authored 10 months ago
  
  9598bb22
- resetting metrics after use · c7f609e1
  Leonard Wayne Hackel authored 10 months ago
  
  c7f609e1
- using metr calc again · fc23d60d
  Leonard Wayne Hackel authored 10 months ago
  
  fc23d60d
- reduce set size · 610133d6
  Leonard Wayne Hackel authored 10 months ago
  
  610133d6
- use .cpu() · cf8819cb
  Leonard Wayne Hackel authored 10 months ago
  
  cf8819cb
- go back to fake data · b4917390
  Leonard Wayne Hackel authored 10 months ago
  
  b4917390
- go back to real data · 75cb8d63
  Leonard Wayne Hackel authored 10 months ago
  
  75cb8d63
- fake metric calculation · 02d5eb43
  Leonard Wayne Hackel authored 10 months ago
  
  02d5eb43
- going back to torchmetrics 1.4 but detaching tensors in val step · 86120b6d
  Leonard Wayne Hackel authored 10 months ago
  
  86120b6d
- changing torchmetric version · 906efd57
  Leonard Wayne Hackel authored 10 months ago
  
  906efd57
- putting mixins back, printing vram in epoch end · c6ae7228
  Leonard Wayne Hackel authored 10 months ago
  
  c6ae7228
- going back to cuda 12, removing hf mixins · 31664f3a
  Leonard Wayne Hackel authored 10 months ago
  
  31664f3a
- using cuda 11 instead · 15bdd384
  Leonard Wayne Hackel authored 10 months ago
  
  15bdd384
- adding max fraction of memory to torch · 3a51a60f
  Leonard Wayne Hackel authored 10 months ago
  
  3a51a60f