Age | Commit message (Collapse) | Author | |
---|---|---|---|
2022-10-30 | Merge master | Muhammad Rizqi Nur | |
2022-10-30 | Merge pull request #3928 from R-N/validate-before-load | AUTOMATIC1111 | |
Optimize training a little | |||
2022-10-30 | Fix dataset still being loaded even when training will be skipped | Muhammad Rizqi Nur | |
2022-10-30 | Add missing info on hypernetwork/embedding model log | Muhammad Rizqi Nur | |
Mentioned here: https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/1528#discussioncomment-3991513 Also group the saving into one | |||
2022-10-30 | Revert "Add cleanup after training" | Muhammad Rizqi Nur | |
This reverts commit 3ce2bfdf95bd5f26d0f6e250e67338ada91980d1. | |||
2022-10-29 | Add cleanup after training | Muhammad Rizqi Nur | |
2022-10-29 | Add input validations before loading dataset for training | Muhammad Rizqi Nur | |
2022-10-29 | Merge branch 'master' into gradient-clipping | Muhammad Rizqi Nur | |
2022-10-29 | Merge branch 'AUTOMATIC1111:master' into 3825-save-hypernet-strength-to-info | timntorres | |
2022-10-29 | Merge pull request #3858 from R-N/log-csv | AUTOMATIC1111 | |
Fix log off by 1 #3847 | |||
2022-10-28 | Fix log off by 1 | Muhammad Rizqi Nur | |
2022-10-28 | Learning rate sched syntax support for grad clipping | Muhammad Rizqi Nur | |
2022-10-28 | Always ignore "None.pt" in the hypernet directory. | timntorres | |
2022-10-28 | Add missing support for linear activation in hypernetwork | benkyoujouzu | |
2022-10-28 | Gradient clipping in train tab | Muhammad Rizqi Nur | |
2022-10-27 | Revert unresolved changes in Bias initialization | AngelBottomless | |
it should be zeros_ or parameterized in future properly. | |||
2022-10-27 | Fix dropout logic | guaneec | |
2022-10-27 | Squashed commit of fixing dropout silently | AngelBottomless | |
fix dropouts for future hypernetworks add kwargs for Hypernetwork class hypernet UI for gradio input add recommended options remove as options revert adding options in ui | |||
2022-10-26 | Fix merge | guaneec | |
2022-10-26 | patch bug (SeverianVoid's comment on 5245c7a) | timntorres | |
2022-10-26 | Merge fix | guaneec | |
2022-10-26 | Merge branch 'master' into hn-activation | guaneec | |
2022-10-26 | Back compatibility | guaneec | |
2022-10-26 | remove duplicate keys and lowercase | AngelBottomless | |
2022-10-26 | Weight initialization and More activation func | AngelBottomless | |
add weight init add weight init option in create_hypernetwork fstringify hypernet info save weight initialization info for further debugging fill bias with zero for He/Xavier initialize LayerNorm with Normal fix loading weight_init | |||
2022-10-26 | Fix off-by-one | guaneec | |
2022-10-26 | Remove activation from final layer of HNs | guaneec | |
2022-10-24 | Removed two unused imports | Melan | |
2022-10-24 | check length for variance | AngelBottomless | |
2022-10-24 | convert deque -> list | AngelBottomless | |
I don't feel this being efficient | |||
2022-10-24 | statistics for pbar | AngelBottomless | |
2022-10-24 | cleanup some code | AngelBottomless | |
2022-10-24 | Hypernetworks - fix KeyError in statistics caching | AngelBottomless | |
Statistics logging has changed to {filename : list[losses]}, so it has to use loss_info[key].pop() | |||
2022-10-23 | Update hypernetwork.py | DepFA | |
2022-10-22 | Allow tracking real-time loss | AngelBottomless | |
Someone had 6000 images in their dataset, and it was shown as 0, which was confusing. This will allow tracking real time dataset-average loss for registered objects. | |||
2022-10-22 | Update hypernetwork.py | AngelBottomless | |
2022-10-22 | small fix | discus0434 | |
2022-10-22 | Merge branch 'AUTOMATIC1111:master' into master | discus0434 | |
2022-10-22 | small fix | discus0434 | |
2022-10-22 | add an option to avoid dying relu | discus0434 | |
2022-10-22 | added a guard for hypernet training that will stop early if weights are ↵ | AUTOMATIC | |
getting no gradients | |||
2022-10-22 | Merge branch 'master' of upstream | discus0434 | |
2022-10-22 | add dropout | discus0434 | |
2022-10-21 | Remove unused variable. | timntorres | |
2022-10-21 | Match hypernet name with filename in all cases. | timntorres | |
2022-10-21 | turns out LayerNorm also has weight and bias and needs to be pre-multiplied ↵ | AUTOMATIC | |
and trained for hypernets | |||
2022-10-21 | Merge branch 'master' into training-help-text | AUTOMATIC1111 | |
2022-10-21 | Revise comments. | timntorres | |
2022-10-21 | Issue #2921-Give PNG info to Hypernet previews. | timntorres | |
2022-10-21 | a more strict check for activation type and a more reasonable check for type ↵ | AUTOMATIC | |
of layer in hypernets |