Kate Cheng comments

Repositories
Issues
Comments

Results 8 comments of


                                            Kate Cheng

[fix] Eagle-2 LLMAPI pybind argument fix.

/bot run

[fix] Eagle-2 LLMAPI pybind argument fix.

/bot kill

[fix] Eagle-2 LLMAPI pybind argument fix.

/bot run --disable-fail-fast

[fix] Eagle-2 LLMAPI pybind argument fix.

/bot kill

[fix] Eagle-2 LLMAPI pybind argument fix.

/bot run --disable-fail-fast

[fix] Eagle-2 LLMAPI pybind argument fix.

/bot run --disable-fail-fast

[fix] Eagle-2 LLMAPI pybind argument fix.

/bot run --disable-fail-fast

policy.train slow at >=32 nodes b/c workers start at different time

@euronymous-aithal I've implemented ray compiled graph and tested on sft algorithm. The original overhead was ~4s for 32 nodes seqlen 48k TP4 CP4 Qwen2.5-14B model. With ray compiled graph, the...