 Eric Wang
|
7fb06c6c22
Revert "Mask out prompt tokens for real"
|
3 lat temu |
|
Eric Wang
|
2204a71505
set EPOCHS back to 3
|
3 lat temu |
|
Eric Wang
|
4a712d4d8e
Mask out prompt tokens for real
|
3 lat temu |
|
Eric Wang
|
fac53721a2
masking bugfix
|
3 lat temu |
 Kohaku-Blueleaf
|
b5a1a0bca7
Add support for valid set size 0 (#83)
|
3 lat temu |
|
Kohaku-Blueleaf
|
0af44f0262
Add option for output dir (#84)
|
3 lat temu |
|
Kohaku-Blueleaf
|
450206caaf
Fix torch.compile call on windows (#81)
|
3 lat temu |
|
Eric Wang
|
cfad895aa1
mask prompt in loss
|
3 lat temu |
 Kakigōri Maker
|
9dab7ba438
add multi-gpu support (ddp) (#54)
|
3 lat temu |
|
Eric Wang
|
f7044049ab
dataset cleaning, visualizations
|
3 lat temu |
|
Eric Wang
|
35029da078
Validation set
|
3 lat temu |
|
Eric Wang
|
5f6614e6fc
Catch outdated installs
|
3 lat temu |
 andreas.echavez
|
1862976b33
Update alpaca-lora to use transformers main branch
|
3 lat temu |
|
Eric Wang
|
2fa1c66388
repair tokenization logic, again
|
3 lat temu |
|
Eric Wang
|
024dde7dab
Revert "fix <eos> tokenization"
|
3 lat temu |
|
Eric Wang
|
6b69ea8665
fix <eos> tokenization
|
3 lat temu |
|
Eric Wang
|
a2607faff0
fix finetuning code :(
|
3 lat temu |
|
Eric Wang
|
d714a73e8c
Update README.md with new checkpoint details
|
3 lat temu |
|
Eric Wang
|
ec98533876
Update README.md; clean up hyperparameters
|
3 lat temu |
|
Eric Wang
|
46ddd2ca85
Ready to go
|
3 lat temu |
|
Eric Wang
|
648af26073
update hyperparams
|
3 lat temu |
|
Eric Wang
|
5cd474bcc0
lr=2e-5
|
3 lat temu |
 Jan Malte Lichtenberg
|
a3b80fdbd5
Fix bug in generate promp using 'instruction' instead of 'input'
|
3 lat temu |
|
Eric Wang
|
41e0ff6c78
tokenizer changes
|
3 lat temu |
|
Eric Wang
|
df2a5dc4be
cleanup notebooks
|
3 lat temu |
|
Eric Wang
|
357ec81a17
decapoda
|
3 lat temu |
|
Eric Wang
|
63121244c8
Licenses and whatnot
|
3 lat temu |
|
Eric Wang
|
26f64780ad
initial commit
|
3 lat temu |
