ml-ryanlee/anchorfix-base-1e18-d896-seed46

base-1e18-d896-seed46

16 effective layers, trained at a 1e18 FLOP budget at this architecture's compute-optimal width. One of a six-seed set (seeds 42-47) in which only the training data order varies; model initialisation is fixed across seeds.

field	value
architecture	base
d_model	896
d_ff	2432
width_ratio	7.0
base_d_model	128
base_d_ff	384
data seed	46
init seed	42

Loads with trust_remote_code=True. Trained with context length 1024; set max_length=1024 when evaluating.