psychedelicious
|
ff72315db2
|
feat(nodes): update SAM backend and nodes to work with SAM points
|
2024-10-23 16:01:15 +11:00 |
|
psychedelicious
|
ba09c1277f
|
feat(nodes): hacked together nodes for segment anything w/ points
|
2024-10-23 16:01:15 +11:00 |
|
David Burnett
|
9bd17ea02f
|
Get flux working with MPS on 2.4.1, with GGUF support
|
2024-10-23 10:20:42 +11:00 |
|
David Burnett
|
24f9b46fbc
|
ruff fix
|
2024-10-23 10:09:24 +11:00 |
|
David Burnett
|
54b3aa1d01
|
load t5 model in the same format as it is saved, seems to load as float32 on Macs
|
2024-10-23 10:09:24 +11:00 |
|
Ryan Dick
|
5cbe89afdd
|
Merge branch 'main' into ryan/flux-ip-adapter-cfg-2
|
2024-10-22 21:17:36 +00:00 |
|
Ryan Dick
|
d20b894a61
|
Add cfg_scale_start_step and cfg_scale_end_step to FLUX Denoise node.
|
2024-10-23 07:59:48 +11:00 |
|
Ryan Dick
|
20362448b9
|
Make negative_text_conditioning nullable on FLUX Denoise invocation.
|
2024-10-23 07:59:48 +11:00 |
|
Ryan Dick
|
5df10cc494
|
Add support for cfg_scale list on FLUX Denoise node.
|
2024-10-23 07:59:48 +11:00 |
|
Ryan Dick
|
da171114ea
|
Naive implementation of CFG for FLUX.
|
2024-10-23 07:59:48 +11:00 |
|
Ryan Dick
|
e545f18a45
|
(minor) Fix ruff.
|
2024-10-21 22:38:06 +00:00 |
|
Ryan Dick
|
e8cd1bb3d8
|
Add FLUX IP-Adapter starter models.
|
2024-10-21 22:17:42 +00:00 |
|
Ryan Dick
|
90a906e203
|
Simplify handling of CLIP ViT selection for FLUX IP-Adapter invocation.
|
2024-10-21 19:54:59 +00:00 |
|
Ryan Dick
|
dde54740c5
|
Test out IP-Adapter with CFG.
|
2024-10-21 15:47:17 +00:00 |
|
Ryan Dick
|
f70a8e2c1a
|
A bunch of HACKS to get ViT-L CLIP vision encoder working for FLUX IP-Adapter. Need to revisit how to clean this all up long term.
|
2024-10-21 15:43:00 +00:00 |
|
Ryan Dick
|
fdccdd52d5
|
Fixes to get XLabsIpAdapterExtension running.
|
2024-10-21 15:43:00 +00:00 |
|
Ryan Dick
|
31ffd73423
|
Initial draft of integrating FLUX IP-Adapter inference support.
|
2024-10-21 15:42:56 +00:00 |
|
Ryan Dick
|
3fa1012879
|
Add IPAdapterDoubleBlocks wrapper to tidy FLUX ip-adapter handling.
|
2024-10-21 15:38:50 +00:00 |
|
Ryan Dick
|
c2a8fbd8d6
|
(minor) Move infer_xlabs_ip_adapter_params_from_state_dict(...) to state_dict_utils.py.
|
2024-10-21 15:38:50 +00:00 |
|
Ryan Dick
|
d6643d7263
|
Add model loading code for xlabs FLUX IP-Adapter (not tested).
|
2024-10-21 15:38:50 +00:00 |
|
Ryan Dick
|
412e79d8e6
|
Add model probing for XLabs FLUX IP-Adapter.
|
2024-10-21 15:38:50 +00:00 |
|
Ryan Dick
|
f939dbdc33
|
Add is_state_dict_xlabs_ip_adapter() utility function.
|
2024-10-21 15:38:50 +00:00 |
|
Ryan Dick
|
24a0ca86f5
|
Add logic for loading an Xlabs IP-Adapter from a state dict.
|
2024-10-21 15:38:50 +00:00 |
|
Ryan Dick
|
95c30f6a8b
|
Add initial logic for inferring FLUX IP-Adapter params from a state_dict.
|
2024-10-21 15:38:50 +00:00 |
|
Ryan Dick
|
ac7441e606
|
Fixup typing/imports for IPDoubleStreamBlockProcessor.
|
2024-10-21 15:38:50 +00:00 |
|
Ryan Dick
|
9c9af312fe
|
Copy IPDoubleStreamBlockProcessor from 47495425db/src/flux/modules/layers.py (L221).
|
2024-10-21 15:38:50 +00:00 |
|
Ryan Dick
|
32c7cdd856
|
Add cfg_scale_start_step and cfg_scale_end_step to FLUX Denoise node.
|
2024-10-21 14:52:02 +00:00 |
|
Mary Hipp
|
bbd89d54b4
|
add it to list
|
2024-10-19 14:08:49 +11:00 |
|
Mary Hipp
|
ee61006a49
|
add starter model
|
2024-10-19 14:08:49 +11:00 |
|
Ryan Dick
|
6df4ee5fc8
|
Make negative_text_conditioning nullable on FLUX Denoise invocation.
|
2024-10-18 20:31:27 +00:00 |
|
Ryan Dick
|
371742d8f9
|
Add support for cfg_scale list on FLUX Denoise node.
|
2024-10-18 20:14:47 +00:00 |
|
Ryan Dick
|
73d4c4d56d
|
Naive implementation of CFG for FLUX.
|
2024-10-16 16:22:35 +00:00 |
|
Mary Hipp
|
5bd87ca89b
|
feat(ui,api): add starter bundles to MM
|
2024-10-15 15:58:11 -04:00 |
|
Ryan Dick
|
ac08c31fbc
|
Remove unnecessary hasattr checks for scaled_dot_product_attention. We pin the torch version, so there should be no concern that this function does not exist.
|
2024-10-10 19:23:45 -04:00 |
|
Ryan Dick
|
ea54a2655a
|
Add a workaround for broken sliced attention on MPS with torch 2.4.1.
|
2024-10-10 19:23:45 -04:00 |
|
Ryan Dick
|
bb6d073828
|
Use the Shakker-Labs ControlNet union model as the only FLUX ControlNet starter model.
|
2024-10-10 13:59:59 +00:00 |
|
Ryan Dick
|
f37c5011f4
|
Reduce peak memory utilization when preparing FLUX controlnet inputs.
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
a654dad20f
|
Remove instantx_control_mode from FLUX ControlNet node.
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
e7f9086006
|
Fix bug with InstantX input image range.
|
2024-10-10 07:59:29 -04:00 |
|
maryhipp
|
5fd3c39431
|
update prepreprocessor logic to be more resilient
|
2024-10-10 07:59:29 -04:00 |
|
maryhipp
|
7daf3b7d4a
|
update starter models to include FLUX controlnets
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
908f65698d
|
Fix support for InstantX non-union models (with no single blocks).
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
63c4ac58e9
|
Support installing InstantX ControlNet models from diffusers directory format.
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
812940eb95
|
(minor) Add comment about future memory optimization.
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
0559480dd6
|
Shift the controlnet-type-specific logic into the specific ControlNet extensions and make the FLUX model controlnet-type-agnostic.
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
d99e7dd4e4
|
Add instantx_control_mode param to FLUX ControlNet invocation.
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
e854181417
|
Create a dedicated FLUX ControlNet invocation.
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
de414c09fd
|
Bugfixes to get InstantX ControlNet working.
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
ce4624f72b
|
Update ControlNetCheckpointProbe.get_base_type() to work with InstantX.
|
2024-10-10 07:59:29 -04:00 |
|
Ryan Dick
|
47c7df3476
|
Fix circular imports related to XLabsControlNetFluxOutput and InstantXControlNetFluxOutput.
|
2024-10-10 07:59:29 -04:00 |
|