Commit History

v5: the verify entrypoint — full logits + activations for the assistant drafter
d25ffc2
verified

HiramKHackenbacker commited on

v4: head-sidecar E4B — fp16 head as a bound input beside the PLE pair
c00efb9
verified

HiramKHackenbacker commited on

card: levered2/ — gather-first embedding, int4 PLE, fp16 head
99dc7f4
verified

HiramKHackenbacker commited on

levered2/ — gather-first embedding, int4 PLE, fp16 head
af07da3
verified

HiramKHackenbacker commited on

levered2/ — gather-first embedding, int4 PLE, fp16 head
35b3812
verified

HiramKHackenbacker commited on

e4b cap bundle: README with provenance, KV arithmetic, digests
0ec8dd0
verified

HiramKHackenbacker commited on

E4B pf64 at cap 131,072 — the checkpoint's own max_position_embeddings
eb11131
verified

HiramKHackenbacker commited on

model card: standard structure, measured facts, verified license declarations
0505916
verified

HiramKHackenbacker commited on

README: ring3 per-layer-type SDPA bundles
5f17787
verified

HiramKHackenbacker commited on

Add ring3 gemma4_e4b_qat_decode_int4lin_tbl_pf64_ring_c4096_gqafull
5cd567c
verified

HiramKHackenbacker commited on

Add ring gemma4_e4b_qat_decode_int4lin_tbl_pf64_ring_c16384
b21d442
verified

HiramKHackenbacker commited on

docs: stable/ measured — memory fix confirmed, capacity-16384 decode cost makes it non-deployable as shipped
366268d
verified

HiramKHackenbacker commited on

prune: remove superseded/non-functional artifacts
d6c45e6
verified

HiramKHackenbacker commited on

prune: remove superseded/non-functional artifacts
963aa31
verified

HiramKHackenbacker commited on

docs: stable/ folders in Status, oracle verification line, known-issue points at available fix
cbf6fbc
verified

HiramKHackenbacker commited on

Add stable gemma4_e4b_qat_decode_int4lin_tbl_pf64_stable_c16384
8e241a6
verified

HiramKHackenbacker commited on

docs: correct known-issue mechanism (shape-signature specialization cache, guided affected equally at depth)
a9385b5
verified

HiramKHackenbacker commited on

docs: refresh measured performance (local retest 2026-08-18)
f9efe59
verified

HiramKHackenbacker commited on

docs: descriptive model card
451139d
verified

Ben Graville commited on

README: w4a8 bundles
36db1ae
verified

Ben Graville commited on

Add w4a8/gemma4_e4b_qat_decode_int4lin_a8_tbl_pf64
41351c7
verified

Ben Graville commited on

add missing chat_template.jinja to gpu-pipelined/gemma4_e4b_qat_decode_int4lin_tbl_pf64/tokenizer (additive fix, bundle otherwise untouched)
9353141
verified

Ben Graville commited on

add missing chat_template.jinja to gpu-pipelined/gemma4_e4b_qat_decode_int4lin/tokenizer (additive fix, bundle otherwise untouched)
cff5afb
verified

Ben Graville commited on

Add PLE gather_raw table sidecar
b83a4d3
verified

Ben Graville commited on

README: ctx16384 bundles
8947bf3
verified

Ben Graville commited on

Add ctx16384 gemma4_e4b_qat_decode_int4lin_tbl_pf64
f847ffa
verified

Ben Graville commited on

README: ctx8192 bundles
18a972a
verified

Ben Graville commited on

Add ctx8192 gemma4_e4b_qat_decode_int4lin_tbl_pf64
642c556
verified

Ben Graville commited on

Add gemma4_e4b_qat_decode_int4lin_tbl_pf64
3dd6670
verified

Ben Graville commited on

Upload README.md with huggingface_hub
65f4cdf
verified

Ben Graville commited on

Add gemma4_e4b_qat_decode_int4lin
f433598
verified

Ben Graville commited on

Upload README.md with huggingface_hub
d85732c
verified

Ben Graville commited on

initial commit
7051b5d
verified

Ben Graville commited on