Update README.md (#58)
c49e251
unverified
Srinivasan Iyer
commited on
Spaces:
Running
on
Zero
Running
on
Zero
Commit History
Fix multiprocessing dataloader checkpointing and use it in the train script (#50)
8c61ab5
unverified
Test first batch matches (#53)
85c2f28
unverified
disable reshard after forward (#56)
9d907fe
unverified
make sure max_encoder_seq_length matches (#55)
48e4ad0
unverified
fix save and reload model state (#49)
22c7fe1
unverified
Add bpb and n_bytes to metric logging (#41)
fe45f69
unverified
Fix init and repro (#48)
aebdc48
unverified
Allow ArrowIterator to read from json (#45)
936d943
unverified
Update checkpointing to use fsspec (#39)
afedb16
unverified
Add rope fp32 (#43)
739dc71
unverified
fix stool (#44)
6fbaf72
unverified
Fix wandb logging (#42)
7cf8fab
unverified
Fix distributed all reduce grad norm (#40)
c79b1fd
unverified
This includes fixes that make checkpointing and reloading work correctly. (#35)
7044771
unverified
Initial codes and scripts for training entropy model (#34)
7622d28
unverified
Use load_async flag to not start MP iterator (#33)
a809259
unverified
Update file check script to check sizes (#32)
bc42ceb
unverified
Fix realtime entropy patching (#26)
392117b
unverified
Ink
commited on
Changes for training entropy model and correcting attention in local models (#25)
6ffeb66
unverified
allow flex-attention to be disabled (#19)
caec8d2
unverified
Ink
commited on
Update preprocess_entropies script to blt inference + add fsspec support (#23)
1da3dd9
unverified
Replace regular filesystem calls with fsspec + add s3 support (#18)
b0120da
unverified
Add plotting code from paper (#17)
d4ddb95
unverified
Package `bytelatent` as a module (#7)
2fdc6f3
unverified
Ink
commited on
