Bigscience176b jz #1

younesbelkada wants to merge 50 commits into bigscience176b from bigscience176b-jz
younesbelkada
younesbelkada small commit
7a284299
younesbelkada first test
50ccdaa1
younesbelkada fix minor issues
ef48867f
younesbelkada minor changes
e3748170
younesbelkada minor fix
ebe52d6c
younesbelkada try smaller model
5baecdae
younesbelkada rm break statement
56a23752
younesbelkada adding new file
682fd3e0
younesbelkada add new file
a0be1eb8
younesbelkada do not load tokenizer
5b1ce6b1
younesbelkada fix device map
9ab38cd2
younesbelkada fix input issue
7418753e
younesbelkada fix logits issue
72e4ca12
younesbelkada new modifs
d495ba99
younesbelkada add print shape
e100f24a
younesbelkada test without att mask
61273fc9
younesbelkada add correct att mask
5ff05870
younesbelkada fix torch.tensor
332e9a4e
younesbelkada add new logits
0ec9f0c6
younesbelkada add few prints
cef597c6
younesbelkada correct print
a254d548
younesbelkada fix print
85351989
younesbelkada add print w&b
d990860b
younesbelkada add .item() to print
cf3f67eb
younesbelkada add some prints
b0edc918
younesbelkada add argmax
295beb21
younesbelkada add new prints
830eb307
younesbelkada put f.linear instead
e5f5e186
younesbelkada remove dropout
d7bcc7b4
younesbelkada add scaled softmax
3687252f
younesbelkada add new arg in config
0b3f4d57
younesbelkada fix import issue
f6339d1e
younesbelkada fix arg issue
9d312d20
younesbelkada fix bias issue
79fe5db3
younesbelkada print shape weight
9e0014cf
younesbelkada print input shape
e15c9c13
younesbelkada add mean bias
a9768dd2
younesbelkada check on bias
292b96ba
younesbelkada add some prints
20acdaa5
younesbelkada change config
89c03c69
younesbelkada add way more prints
fe6de750
younesbelkada add new prints
84565e6e
younesbelkada add if statement
f95e1c57
younesbelkada add scale softmax
73218312
younesbelkada save logits
0c3d92b0
younesbelkada final changes
dfd3b845
younesbelkada add apex func
d508f508
younesbelkada add new tensors
c795adad
younesbelkada add .to(dtype)
eede2a3b
younesbelkada add message
37f869b0
younesbelkada younesbelkada closed this 4 years ago

Login to write a write a comment.

Login via GitHub

Reviewers
No reviews
Assignees
No one assigned
Labels
Milestone