aboutsummaryrefslogtreecommitdiffstats
path: root/modules/sd_models.py
Commit message (Collapse)AuthorAgeFilesLines
* Fix forced reloadKohaku-Blueleaf2023-12-061-1/+1
|
* Ensure the cached weight will not be affectedKohaku-Blueleaf2023-12-021-2/+2
|
* Merge branch 'dev' into test-fp8Kohaku-Blueleaf2023-12-021-4/+13
|\
| * Add support for SD 2.1 Turbo, by converting the state dict from SGM to LDM ↵MrCheeze2023-12-021-4/+13
| | | | | | | | on load
* | Fix pre-fp8Kohaku-Blueleaf2023-11-251-1/+1
| |
* | Option for using fp16 weight when apply loraKohaku-Blueleaf2023-11-211-3/+11
| |
* | Use options instead of cmd_argsKohaku-Blueleaf2023-11-191-29/+32
| |
* | Merge branch 'dev' into test-fp8Kohaku-Blueleaf2023-11-161-1/+4
|\|
| * more changes for #13865: fix formatting, rename the function, add comment ↵AUTOMATIC11112023-11-051-1/+1
| | | | | | | | and add a readme entry
| * linterAUTOMATIC11112023-11-051-2/+2
| |
| * Merge branch 'dev' into masterAUTOMATIC11112023-11-051-20/+30
| |\
| * | Use devices.torch_gc() instead of empty_cache()Ritesh Gangnani2023-11-051-1/+0
| | |
| * | Add SSD-1B as a supported modelRitesh Gangnani2023-11-051-2/+6
| | |
* | | ManualCast for 10/16 series gpuKohaku-Blueleaf2023-10-281-9/+12
| | |
* | | ignore mps for fp8Kohaku-Blueleaf2023-10-251-1/+3
| | |
* | | Fix alphas cumprodKohaku-Blueleaf2023-10-251-1/+2
| | |
* | | Fix alphas_cumprod dtypeKohaku-Blueleaf2023-10-251-0/+1
| | |
* | | fp8 for TEKohaku-Blueleaf2023-10-251-0/+7
| | |
* | | Fix lintKohaku-Blueleaf2023-10-231-1/+1
| | |
* | | Add CPU fp8 supportKohaku-Blueleaf2023-10-231-4/+16
| | | | | | | | | | | | | | | | | | Since norm layer need fp32, I only convert the linear operation layer(conv2d/linear) And TE have some pytorch function not support bf16 amp in CPU. I add a condition to indicate if the autocast is for unet.
* | | Add sdxl only argKohaku-Blueleaf2023-10-191-0/+3
| | |
* | | Add fp8 for sd unetKohaku-Blueleaf2023-10-191-0/+3
| |/ |/|
* | repair unload sd checkpoint buttonAUTOMATIC11112023-10-151-12/+1
| |
* | use shallow copy for #13535AUTOMATIC11112023-10-141-2/+1
| |
* | Merge pull request #13535 from chu8129/devAUTOMATIC11112023-10-141-4/+5
|\ \ | | | | | | fix: checkpoints_loaded:{checkpoint:state_dict}, model.load_state_dict issue in dict value empty
| * | reverstwangqiuwen2023-10-071-0/+2
| | |
| * | upwangqiuwen2023-10-071-6/+5
| | |
* | | Merge pull request #13139 from AUTOMATIC1111/ckpt-dir-path-separatorAUTOMATIC11112023-09-301-2/+3
|\ \ \ | | | | | | | | fix `--ckpt-dir` path separator and option use `short name` for checkpoint dropdown
| * | | parsing string to pathw-e-w2023-09-081-2/+3
| | | |
| * | | keep order in list of checkpoints when loading model that doesn't have a ↵AUTOMATIC11112023-08-301-1/+21
| | | | | | | | | | | | | | | | checksum
* | | | add missing import, simplify code, use patches module for #13276AUTOMATIC11112023-09-301-7/+12
| | | |
* | | | Merge pull request #13276 from woweenie/patch-1AUTOMATIC11112023-09-301-1/+14
|\ \ \ \ | |_|/ / |/| | | patch DDPM.register_betas so that users can put given_betas in model yaml
| * | | patch DDPM.register_betas so that users can put given_betas in model yamlwoweenie2023-09-151-1/+14
| | |/ | |/|
* | | fix王秋文/qwwang2023-09-181-0/+1
| | |
* | | use dict[key]=model; did not update orderdict order, should use move to endqiuwen.wang2023-09-151-0/+1
|/ /
* / keep order in list of checkpoints when loading model that doesn't have a ↵AUTOMATIC11112023-08-301-1/+21
|/ | | | checksum
* set devices.dtype_unet correctlyAUTOMATIC11112023-08-231-1/+2
|
* add --medvram-sdxlAUTOMATIC11112023-08-221-8/+8
|
* Fix for consistency with shared.opts.sd_vae of UIUminosachi2023-08-211-0/+1
|
* Change where VAE state are stored in modelUminosachi2023-08-201-2/+0
|
* Change to access sd_model attribute with dotUminosachi2023-08-201-2/+2
|
* Store base_vae and loaded_vae_file in sd_modelUminosachi2023-08-201-16/+8
|
* Fix SD VAE switch error after model reuseUminosachi2023-08-201-2/+20
|
* resolve the issue with loading fp16 checkpoints while using --no-halfAUTOMATIC11112023-08-171-1/+4
|
* send weights to target device instead of CPU memoryAUTOMATIC11112023-08-161-1/+16
|
* Revert "send weights to target device instead of CPU memory"AUTOMATIC11112023-08-161-1/+1
| | | | This reverts commit 0815c45bcdec0a2e5c60bdd5b33d95813d799c01.
* send weights to target device instead of CPU memoryAUTOMATIC11112023-08-161-1/+1
|
* put refiner into main UI, into the new accordions sectionAUTOMATIC11112023-08-121-0/+3
| | | | | | add VAE from main model into infotext, not from refiner model option to make scripts UI without gr.Group fix inconsistencies with refiner when usings samplers that do more denoising than steps
* resolve merge issuesAUTOMATIC11112023-08-101-2/+5
|
* Merge branch 'dev' into refinerAUTOMATIC11112023-08-101-11/+12
|\