transformers
b7601f73 - Keep MiMo-V2-Flash's router in float32 in bfloat16 models (#50722)

Commit
22 hours ago
Keep MiMo-V2-Flash's router in float32 in bfloat16 models (#50722) * float32 router * Declare MiMo-V2-Flash's router in float32 * remove dtype change * remove test
Parents
Loading