leejet
27887b630d
fix ci error
2024-03-03 14:52:08 +08:00
bssrdf
70c3397b60
remove MultiheadAttention2; customized MultiheadAttention
2024-02-28 11:04:03 -05:00
bssrdf
fd098afc6d
bring back freeing pmid lora params buffer; simply pooled output of CLIPvision
2024-02-27 14:49:39 -05:00
bssrdf
7b72e17b7e
a bit clean up
2024-02-26 16:34:04 -05:00
bssrdf
77b6a42bc3
photomaker working properly now after merging and adapting to GGMLBlock API
2024-02-26 13:32:19 -05:00
bssrdf
b556d4411c
fixed bugs; now failed with cuda error; likely out-of-mem on GPU
2024-02-25 22:55:26 -05:00
bssrdf
d35861b501
merge with a master again; build ok
2024-02-25 19:30:22 -05:00
bssrdf
53beb92d2f
merged with master and resolved conflicts; adapted to GGMLBlock API
2024-02-25 14:35:37 -05:00
leejet
4a8190405a
fix: fix the issue with dynamic linking
2024-02-25 21:39:01 +08:00
leejet
730585d515
sync: update ggml ( #180 )
2024-02-25 21:11:01 +08:00
leejet
b6368868d9
feat: introduce GGMLBlock and implement SVD(Broken) ( #159 )
...
* introduce GGMLBlock and implement SVD(Broken)
* add sdxl vae warning
2024-02-24 20:06:39 +08:00
bssrdf
8b89e0ea6b
added an option for normalizing input ID images; cleaned up debugging code
2024-02-23 10:42:45 -05:00
bssrdf
a6a676b646
add style_ratio as a cli param; reworked encode with trigger for attention weights
2024-02-19 12:49:41 -05:00
bssrdf
d38b85d87e
fixed a bug in ggml's conv_2d, and now > 1 input images working
2024-02-18 22:18:53 -05:00
bssrdf
6064b0e28b
fixed a decode bug
2024-02-16 14:17:00 -05:00
bssrdf
62b0a9b9f8
turn pmid lora apply back on
2024-02-16 11:50:46 -05:00
bssrdf
191c5ec05a
fixed a bug in FuseBlock forward; also remove diag_mask op in for vision transformer; getting better results
2024-02-15 18:57:19 -05:00
bssrdf
26c5cca043
small tweak; still not getting ID faces
2024-02-14 21:22:40 -05:00
bssrdf
539a94adc7
add a few print tensor; tweak in sample again
2024-02-13 21:44:17 -05:00
bssrdf
b0579ec460
added photomaker lora model (to be tested)
2024-02-12 16:31:02 -05:00
bssrdf
857af489a9
fixed stat_merge_step
2024-02-11 22:32:35 -05:00
bssrdf
5a13b48b6e
fixed wtype issue; input images can only be 1 because issue with transformer when batch size > 1 (to be investigated)
2024-02-11 12:54:18 -05:00
bssrdf
807c340118
more progress; now alloc buffer failed
2024-02-10 22:53:54 -05:00
bssrdf
d93951456d
debuging graph compute
2024-02-09 22:21:04 -05:00
bssrdf
304704fa36
corrected clip vision model output
2024-02-08 22:16:01 -05:00
bssrdf
0a38d84dcc
debugging vision_model outputs
2024-02-08 17:27:26 -05:00
bssrdf
7f5f580532
making more progress; finishing vision model forward
2024-02-07 17:25:06 -05:00
bssrdf
87fcee0851
making progress; need to implement tokenizer decoder
2024-02-06 22:17:40 -05:00
bssrdf
ad7ec45ec1
add a convert_token_to_id function for photomaker to extract trigger word's token id
2024-02-06 16:17:07 -05:00
bssrdf
df7f64243b
add a get_learned_condition_with_trigger function to do photomaker stuff
2024-02-06 16:30:09 -05:00
bssrdf
78651f2535
fixed soem bugs; now photomaker model weights can be loaded into their tensor buffers
2024-02-04 17:56:18 -05:00
bssrdf
cbfa702fad
added PhotoMakerIDEncoder model in SD
2024-02-04 16:03:00 -05:00
bssrdf
5dbbea0ca4
first efforts at implementing photomaker; lots more to do
2024-02-02 18:36:40 -05:00
leejet
349439f239
style: format code
2024-01-29 23:05:18 +08:00
Steward Garcia and leejet
36ec16ac99
feat: Control Net support + Textual Inversion (embeddings) ( #131 )
...
* add controlnet to pipeline
* add cli params
* control strength cli param
* cli param keep controlnet in cpu
* add Textual Inversion
* add canny preprocessor
* refactor: change ggml_type_sizef to ggml_row_size
* process hint once time
* ignore the embedding name case
---------
Co-authored-by: leejet <leejet714@gmail.com >
2024-01-29 22:38:51 +08:00
leejet
2b6ec97fe2
sync: update ggml ( #134 )
2024-01-05 23:18:41 +08:00
leejet
2e79a82f85
refactor: reorganize code and use c api ( #133 )
2024-01-01 16:22:18 +08:00
leejet