YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
elicit-A1-tablefreeze-e2
Run C of the terminator debug, second epoch: the token tables trained for one epoch and were then frozen while the linear layers continued. Failed. Acts in 10% of acceptance samples, and that figure is an upper bound because the model stopped in only 6.7% of samples, so most ran to the token cap.
Qwen2.5-32B base never trained the ChatML control tokens. <|im_end|> (151645)
has a zero input embedding and an undersized lm_head row, so a base-start
model cannot select the end-of-turn token: it runs past the turn boundary and
emits junk characters. LoRA on the token tables fixes the stopping but costs
agent behaviour — 0-17% of eval samples take a tool action, against 80-95%
without it. These arms are the search for a recipe that keeps both.