Hailo-10H / DFC v5.3.0: a16_w16 on a transformer encoder is not a blanket allocator wall — it's a 3-stage cascade (attention crash → a16-conv nan exponent → needs super-defuse). What is the intended 16-bit path?

Hi Fernando — this is genuinely useful, thanks for going through the repo and reporting back in such detail. On your direct question : no, I haven’t gotten a16 through the attention block either. My own a16_w16 usage is limited to input_layer1 (the embedding input, far from attention — matches what you saw) and ew_add* layers on the encoder recipe.