Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 17:24:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

Update changelog

Browse Source
This commit is contained in:
Disty0
2025-01-31 20:27:22 +03:00
parent 95060134b9
commit d2fee97e24
2 changed files with 5 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CHANGELOG.md
+4 -3
View File
@@ -9,7 +9,8 @@
- **Docs**:
- Updated [Debugging guide](https://github.com/vladmandic/automatic/wiki/Debug)
- **Torch**:
- for cuda environemnts set default to `torch==2.6.0+cu126`
- for cuda environments set default to `torch==2.6.0+cu126`
for rocm environments set default to `torch==2.6.0+rocm6.2.4`
*note*: to avoid disruptions sdnext does not perform torch install during in-place upgrades
to force torch upgrade, either start with new installation or use `--reinstall` flag
- add support for torch **tunable ops**, this can speed up operations by up to *10-30%* on some platforms
@@ -25,12 +26,12 @@
- detailer with faceid modules
- detailer restore pipeline before run
- fix `python==3.9` compatibility
- handle invalid `triton` with `torch==2.6.0`
- handle invalid `triton` on Linux
- correct library import order
- update requirements
- calculate dyn atten bmm slice rate
- dwpose update and patch `mmengine` installer
- wrap extensions using incorrect device
- ipex device wrapper with adetailer
- openvino error handling
- relax python version checks for rocm
modules/devices.py
+1 -1
View File
@@ -307,7 +307,7 @@ def set_cuda_tunable():
pass
torch.cuda.tunable.set_filename(fn)
if torch.cuda.tunable.is_enabled():
log.debug(f'Torche tunable: enabled={torch.cuda.tunable.is_enabled()} tuning={torch.cuda.tunable.tuning_is_enabled()} iterations={torch.cuda.tunable.get_max_tuning_iterations()} duration={torch.cuda.tunable.get_max_tuning_duration()} fn="{fn}" entries={lines}')
log.debug(f'Torch tunable: enabled={torch.cuda.tunable.is_enabled()} tuning={torch.cuda.tunable.tuning_is_enabled()} iterations={torch.cuda.tunable.get_max_tuning_iterations()} duration={torch.cuda.tunable.get_max_tuning_duration()} fn="{fn}" entries={lines}')
except Exception as e:
log.warning(f'Torch tunable: {e}')