D1A for PyTorch (CUDA / CPU, non-Mac)

JohnP1 's Collections

updated 3 days ago

D1A decision models as PyTorch checkpoints (LoRA adapter + pointer head on Gemma 4), for CUDA GPUs, CPUs and any non-Apple platform.

JohnP1/d1a-e2b-it-qat

Text Classification • Updated 3 days ago • 49

Note E2B-it QAT v0.6, PyTorch adapter + pointer head on Google's QAT int4 trunk: decisions, chat, photo/voice/video (v0.7 is MLX-only so far)
JohnP1/d1a-e4b-it-qat

Text Classification • Updated 5 days ago • 27

Note E4B-it QAT, PyTorch adapter + pointer head on Google's QAT int4 trunk: decisions, chat, photo/voice/video; level with v0.6 overall, typesafe −7.8
JohnP1/d1a-e4b

Text Classification • Updated 9 days ago • 144

Note E4B, PyTorch adapter + pointer head (v0.6 = latest)
JohnP1/d1a-e4b-it

Text Classification • Updated 9 days ago • 33

Note E4B on the instruction-tuned Gemma 4 trunk: decisions + chat with the adapter off
JohnP1/d1a-e2b

Text Classification • Updated 9 days ago • 120

Note E2B, PyTorch adapter + pointer head
JohnP1/d1a-e2b-it

Text Classification • Updated 9 days ago • 80 • 1

Note E2B on the instruction-tuned Gemma 4 trunk, experimental (for apps that already load E2B-it)