Making large AI models cheaper, faster and more accessible
COMMITS
May 29, 2025
Y
address conversation
YeAnbang committed
May 22, 2025
T
fix default eval setting (#6321)
Tong Li committed
May 28, 2025
Y
address conversation
YeAnbang committed
May 21, 2025
Y
fix missing tags parameter
YeAnbang committed
May 20, 2025
T
fix empty tensor (#6319)
Tong Li committed
Y
add uuid to rollout log
YeAnbang committed
Y
fix metric calculation
YeAnbang committed
May 17, 2025
Y
fix logging rollouts
YeAnbang committed
May 16, 2025
Y
upgrade reward functions
YeAnbang committed
Y
support logging rollouts to wandb
YeAnbang committed
Y
address conversation
YeAnbang committed
Y
fix evaluation
YeAnbang committed
Y
remove redundant code and fix bugs
YeAnbang committed
May 15, 2025
T
handle empty index
Tong Li committed
Y
move prompt-level-filtering to buffer side
YeAnbang committed
Y
move prompt-level-filtering to buffer side
YeAnbang committed
Y
disable wandb tb syncing
YeAnbang committed
Y
use consumer global step
YeAnbang committed
May 14, 2025
T
[feat] Support prompt level dynamic (#6300)
Tong Li committed
Y
move logging to producer
YeAnbang committed
April 30, 2025
Y
Support evaluation during training
YeAnbang committed
Y
upgrade reward math verification
YeAnbang committed
Y
reuse comm-group
YeAnbang committed
Y
Support evaluation during training
YeAnbang committed
T
[feat] Sync shard model (#6289)
Tong Li committed
May 13, 2025
T
update pad seq (#6303)
Tong Li committed
May 7, 2025
Y
[fix] revert reward update and evaluation (#6295)
YeAnbang committed
May 1, 2025
Y
rewrite reward fn
YeAnbang committed
April 29, 2025
Y
[feat] Support boxed math reward (#6284)
YeAnbang committed