prompt fixes 8c4ef5c
div18 commited on
Commit History
prompt changes 8425a53
div18 commited on
prompts c157063
div18 commited on
fix CPU tensors e408dca
div18 commited on
fix bug 69f37e9
div18 commited on
reward etc tuning 67810ba
div18 commited on
env changes 70cdeae
div18 commited on
training changes 7dbb622
div18 commited on
entropy spread d23c9c4
div18 commited on
OOM 0f6141d
div18 commited on
fixes d222529
div18 commited on
fixes 46cd5c4
div18 commited on
fix backprop 871c1ae
div18 commited on
changes d41d25d
div18 commited on
fix structure fa8da3f
div18 commited on
edits 619e74d
div18 commited on
fix: disable Qwen thinking at Jinja template level with enable_thinking=False 5edb1ce
div18 commited on
fix: disable Qwen thinking at Jinja template level with enable_thinking=False 65788cc
div18 commited on
fix: add /no_think to system prompts, strip TRACE blocks from model output 9fd06fa
div18 commited on
fixes 75c8df1
div18 commited on
code e890160
div18 commited on