a524d137d0
patch bug (SeverianVoid's comment on 5245c7a)
3 years ago
7207e3bf49
remove duplicate keys and lowercase
3 years ago
de096d0ce7
Weight initialization and More activation func
...
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
3 years ago
e9a410b535
check length for variance
3 years ago
0d2e1dac40
convert deque -> list
...
I don't feel this being efficient
3 years ago
348f89c8d4
statistics for pbar
3 years ago
40b56c9289
cleanup some code
3 years ago
b297cc3324
Hypernetworks - fix KeyError in statistics caching
...
Statistics logging has changed to {filename : list[losses]}, so it has to use loss_info[key].pop()
3 years ago
1fbfc052eb
Update hypernetwork.py
3 years ago
48dbf99e84
Allow tracking real-time loss
...
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.
3 years ago
24694e5983
Update hypernetwork.py
3 years ago
6a4fa73a38
small fix
3 years ago
97749b7c7d
Merge branch 'AUTOMATIC1111:master' into master
3 years ago
7912acef72
small fix
3 years ago
fccba4729d
add an option to avoid dying relu
3 years ago
7fd90128eb
added a guard for hypernet training that will stop early if weights are getting no gradients
3 years ago
dcb45dfecf
Merge branch 'master' of upstream
3 years ago
0e8ca8e7af
add dropout
3 years ago
272fa527bb
Remove unused variable.
3 years ago
19818f023c
Match hypernet name with filename in all cases.
3 years ago
03a1e288c4
turns out LayerNorm also has weight and bias and needs to be pre-multiplied and trained for hypernets
3 years ago
0c5522ea21
Merge branch 'master' into training-help-text
3 years ago
4ff274e1e3
Revise comments.
3 years ago
5245c7a493
Issue #2921-Give PNG info to Hypernet previews.
3 years ago
c23f666dba
a more strict check for activation type and a more reasonable check for type of layer in hypernets
3 years ago
f89829ec3a
Revert "fix bugs and optimizations"
...
This reverts commit 108be15500
3 years ago
108be15500
fix bugs and optimizations
3 years ago
a71e021236
only linear
3 years ago
d8acd34f66
generalized some functions and option for ignoring first layer
3 years ago
6f98e89486
update
3 years ago
d6ea584137
change html output
3 years ago
2ce52d32e4
fix for #3086 failing to load any previous hypernet
3 years ago
42fbda83bb
layer options moves into create hnet ui
3 years ago
7f8670c4ef
Merge branch 'master' into master
3 years ago
da72becb13
Use training width/height when training hypernetworks.
3 years ago
e40ba281f1
update
3 years ago
a5611ea502
update
3 years ago
6021f7a75f
add options to custom hypernetwork layer structure
3 years ago
703e6d9e4e
check NaN for hypernetwork tuning
3 years ago
c7a86f7fe9
add option to use batch size for training
3 years ago
03d62538ae
remove duplicate code for log loss, add step, make it read from options rather than gradio input
3 years ago
326fe7d44b
Merge remote-tracking branch 'Melanpan/master'
3 years ago
c344ba3b32
add option to read generation params for learning previews from txt2img
3 years ago
354ef0da3b
add hypernetwork multipliers
3 years ago
8636b50aea
Add learn_rate to csv and removed a left-over debug statement
3 years ago
1cfc2a1898
Save a csv containing the loss while training
3 years ago
c3c8eef9fd
train: change filename processing to be more simple and configurable
...
train: make it possible to make text files with prompts
train: rework scheduler so that there's less repeating code in textual inversion and hypernets
train: move epochs setting to options
3 years ago
ee015a1af6
change textual inversion tab to train
...
remake train interface to use tabs
3 years ago
2d006ce16c
xy_grid: Find hypernetwork by closest name
3 years ago
d6fcc6b87b
apply lr schedule to hypernets
3 years ago
timntorres
AngelBottomless
DepFA
discus0434
discus0434
AUTOMATIC
Silent
Melan
Milly