Skip to content

npu-pack: Qwen3-8B 1.9 pack imports end to end (hidden 4096 / 32 heads / 36 layers), head read from the decode If branch #2009

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
bong-water-water-bong merged 20 commits into from
Nov 10, 2026
Merged

npu-pack: Qwen3-8B 1.9 pack imports end to end (hidden 4096 / 32 heads / 36 layers), head read from the decode If branch #2009

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
20 commits
Select commit Hold shift + click to select a range
7c4073c
npu: measure the 1.9 Token fconst rule for hidden 4096 / 32 heads / 3…
Nov 10, 2026
95f1a20
npu-pack: read lm_head from the If branch that is the decode path, so…
Nov 10, 2026
e478d6c
npu-pack: decide the decode If branch by which branch holds the weigh…
Nov 10, 2026
1516de5
npu: the integer arm's operands come from the engine's packer, not a …
Nov 10, 2026
7904dfa
npu: keyed dumps settle it -- the device's weight scales are BIT-IDEN…
Nov 10, 2026
d5f7d48
npu: the integer head, and the gate it unlocks -- device logits BIT-E…
Nov 10, 2026
49fcca1
Merge origin/main into 1bit/host-gemm-i8-fix: resolve the arm's quant…
Nov 10, 2026
297d9ba
npu-pack: read the fused SSMLP/GQO FFN attributes, and the 4096/32/40…
Nov 10, 2026
7e6481e
docs/npu: say what NPU_INFER_HOST_GEMM=i8 is for, and what it measured
Nov 10, 2026
c7c5043
docs/npu: the two-arm command sequence for the bit-exact gate
Nov 10, 2026
172178d
npu-pack: withdraw the 4096/32/40 row -- a prefix cannot be inferred,…
Nov 10, 2026
043e58b
npu-pack: the h4096 36-layer fconst row, so the Qwen3-8B pack imports…
Nov 10, 2026
452975c
npu-pack test: the new 36-layer row goes in BOTH places, as the guard…
Nov 10, 2026
8761fac
Merge remote-tracking branch 'origin/main' into fconst19-4096-36
Nov 10, 2026
62cfd5e
npu-pack: fused-MLP attributes are not graph matmuls, and the 4096/32…
Nov 10, 2026
f03b84a
npu-pack test: arm 2b removes its second fixture directory
Nov 10, 2026
236febb
Merge origin/main into 1bit/fconst19-36layer: keep only the rule row …
Nov 10, 2026
234bb82
Merge branch 'main' into 1bit/fconst19-36layer
bong-water-water-bong Nov 10, 2026
b65e3a2
Merge main into fconst19-4096-36
Nov 10, 2026
fb590f6
Merge #1980's branch (1bit/fconst19-36layer) into fconst19-4096-36
Nov 10, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Loading
Loading