Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
151 commits
Select commit Hold shift + click to select a range
9b81d1f
perf: report native CPU decode profile
justinchuby Jul 22, 2026
4c6712b
docs: compare CPU decode runtimes
justinchuby Jul 22, 2026
58a3324
perf(cpu): rename SQNBit decode knob and route slow-dequant M=1 to MLAS
justinchuby Jul 22, 2026
145549a
perf(cpu-gqa): vectorize decode QK scores and P·V accumulation
justinchuby Jul 22, 2026
2e982c7
perf(cpu): bulk-copy fast path for contiguous f32 kernel I/O
justinchuby Jul 22, 2026
c9762b6
test(cpu-gqa): validate AVX2 dot/axpy paths and correct error bound
justinchuby Jul 22, 2026
32a122e
perf(cpu): NUMA-local CPU affinity for the M=1 decode pool
justinchuby Jul 22, 2026
046414b
perf(cpu): fix NUMA affinity mask sizing, diagnostics, compact select…
justinchuby Jul 22, 2026
96fd406
chore(squad): merge decision inbox for CPU EP perf session
justinchuby Jul 22, 2026
7d78804
perf(cpu): opt-in numa-split two-level decode for M=1 int4
Jul 22, 2026
82e44be
test(cpu-gqa): make SIMD tests portable to non-AVX2 x86 and ARM
justinchuby Jul 22, 2026
f8848c9
feat(cpu): f16/bf16 activations + topology-derived tuning for MatMulN…
justinchuby Jul 22, 2026
122b31a
perf(cpu): cross-OS decode affinity (Windows/macOS) + safe auto-enable
justinchuby Jul 22, 2026
160f4fc
chore(squad): merge generality-batch decisions inbox
justinchuby Jul 22, 2026
eb19730
perf(cpu-decode): persistent SPMD decode pool behind ONNX_GENAI_CPU_D…
justinchuby Jul 22, 2026
cee3c20
fix(cpu-spmd): real parity test + inspectable numa/spmd precedence + …
justinchuby Jul 23, 2026
53fe7c0
chore(squad): merge SPMD decode notes, untrack inbox drop-box
justinchuby Jul 23, 2026
170eb9e
Enable f16 skip simplified layer norm
justinchuby Jul 23, 2026
8df07d9
perf(cpu-matmulnbits): route accuracy_level 0/1 to MLAS SQNBit CompFp32
justinchuby Jul 23, 2026
95876d5
perf(cpu-decode): run GroupQueryAttention on the shared decode pool
justinchuby Jul 23, 2026
770d0c5
perf(cpu-decode): make the persistent SPMD pool the default decode path
justinchuby Jul 23, 2026
b820a87
fix(cpu-decode): defer default-on SPMD to flat path on explicit affinity
justinchuby Jul 23, 2026
536025f
chore(squad): record persistent-SPMD-default decision
justinchuby Jul 23, 2026
eedbf93
perf(cpu-ep): kill per-token KV re-widen in f16 GQA decode (2.5-3.1x)
justinchuby Jul 23, 2026
4caa6d4
chore(squad): record f16-GQA decode + foundry cross-model comparison
justinchuby Jul 23, 2026
8638ec6
test(cpu-ep): bit-exact f16 GQA full-widen parity
justinchuby Jul 23, 2026
80b09c3
test(cpu-ep): independent no-tail-with-past present-cache assertion
justinchuby Jul 23, 2026
20d0dab
chore(squad): record f16 lazy-widen parity test closure
justinchuby Jul 23, 2026
8598f6a
perf(cpu-ep): route f16 int4 MatMulNBits decode GEMV onto SPMD workers
justinchuby Jul 23, 2026
08875b1
test: cover f16 MatMulNBits shard parity
justinchuby Jul 23, 2026
515e34d
chore(squad): record f16 MatMulNBits SPMD-shard win (8598f6a+08875b1)
justinchuby Jul 23, 2026
475fa47
perf(cpu-gqa): bound rotary cos/sin widen to addressed row prefix
justinchuby Jul 23, 2026
6941a9a
test(cpu-gqa): close Chew rotary parity coverage findings
justinchuby Jul 23, 2026
fad338f
chore(squad): record f16 GQA rotary-prefix widen win (475fa47+6941a9a)
justinchuby Jul 23, 2026
96ab5ec
perf(cpu): vectorize SiLU via MLAS logistic and generalize contiguous…
justinchuby Jul 23, 2026
4595551
fix(cpu): make MLAS SiLU robust at extremes/Inf/NaN; strengthen tests
justinchuby Jul 23, 2026
1a2f5a1
chore(squad): record SiLU MLAS robustness remediation (4595551)
justinchuby Jul 23, 2026
316113e
perf(cpu): vectorize skip simplified layernorm
justinchuby Jul 23, 2026
d047106
chore(squad): record SkipLayerNorm SIMD win (316113e)
justinchuby Jul 23, 2026
d14cc83
perf(cpu): vectorize widened f16 SiLU
justinchuby Jul 23, 2026
28adcd9
perf(cpu): fold QKV-bias Add into MatMulNBits GEMV epilogue
justinchuby Jul 23, 2026
7eb9099
chore(squad): record f16 SiLU + QKV-bias fusion wins (d14cc83, 28adcd9)
justinchuby Jul 23, 2026
963427b
test(cpu): assert bias-Add fusion skips already-biased MatMulNBits
justinchuby Jul 23, 2026
ea25b20
chore(squad): record final CPU whole-model benchmark win vs ort-genai
justinchuby Jul 23, 2026
1be1bd5
test(cpu): add cross-dtype SIMD-vs-scalar parity regression gate for …
justinchuby Jul 23, 2026
c878f75
chore(squad): record CPU SIMD parity regression gate
justinchuby Jul 23, 2026
84b40d9
feat(cpu): widen core kernels for bf16
justinchuby Jul 23, 2026
a68b076
feat(cpu): extend bf16 attention and movement coverage
justinchuby Jul 23, 2026
b9b5f45
chore(squad): record bf16 CPU EP coverage landing
justinchuby Jul 23, 2026
2fbc679
perf(engine): cut greedy decode loop overhead
justinchuby Jul 23, 2026
8f55928
feat(bench): add --backend {native,ort,auto} to profile_native for ap…
justinchuby Jul 23, 2026
557c3ed
test(cpu): lock numerical divergence root causes
justinchuby Jul 23, 2026
8244e65
chore(squad): record loop-overhead + backend A/B + divergence root-ca…
justinchuby Jul 23, 2026
d5cd0a8
feat(bench): add profile_vision native-vs-ort A/B for CNN/detector mo…
justinchuby Jul 23, 2026
6604295
feat(cpu): add MLAS-backed Conv and Pool kernels
justinchuby Jul 23, 2026
ab41f43
chore(squad): record CNN Conv/Pool landing + textproto fixture PR
justinchuby Jul 23, 2026
70cd499
fix(cpu): per-block int8 activation quant for MatMulNBits CompInt8
justinchuby Jul 23, 2026
1932aee
test(cpu): drop superseded CompInt8 argmax-reversal characterization …
justinchuby Jul 23, 2026
2c4cfab
fix(cpu): generalize GQA rotary dimensions
justinchuby Jul 23, 2026
fd302e5
feat(cpu): implement CausalConvWithState + LinearAttention (Qwen3.5 g…
justinchuby Jul 23, 2026
d91d776
feat(cpu): enable Qwen3.5 native end-to-end (GatherBlockQuantized, co…
justinchuby Jul 23, 2026
feeb430
docs(squad): record CompInt8/phi3/qwen3.5 merge decisions
justinchuby Jul 23, 2026
272438f
perf(cpu): eliminate Qwen3.5 decode materialization
justinchuby Jul 23, 2026
db188b3
docs(squad): record qwen3.5 decode 180x merge + follow-up nits
justinchuby Jul 23, 2026
780ddbc
feat(mlas-sys): expose MLAS NCHWc blocked-convolution API
justinchuby Jul 23, 2026
9f93d3a
perf(cpu): NCHWc Conv with weight pre-packing + Conv+BN(+Relu) fusion
justinchuby Jul 23, 2026
607e0a6
docs(squad): record NCHWc Conv + fusion merge
justinchuby Jul 23, 2026
c38438e
feat(cpu): support f16 RotaryEmbedding
justinchuby Jul 23, 2026
5e7abc8
docs(squad): record f16 RoPE/Gemma + foundry breadth + generality gaps
justinchuby Jul 23, 2026
bac0ae3
test(cpu): oracle regression for 8-bit block-128 MatMulNBits execute …
justinchuby Jul 23, 2026
a87365a
docs(squad): record token-divergence resolution + 8-bit oracle regres…
justinchuby Jul 23, 2026
05a96bd
perf(cpu-ep): NCHWc graph-level layout propagation for CNN backbones
justinchuby Jul 23, 2026
40acb5f
feat(cpu): support rank-3 1-D Conv via rank-4 NCHW lifting
justinchuby Jul 23, 2026
4c25e43
docs(squad): record NCHWc layout prop (Brass) + rank-3 1-D Conv (Delko)
justinchuby Jul 23, 2026
209a56b
feat(cpu): extend bfloat16 operator coverage
justinchuby Jul 23, 2026
601d2d7
docs(squad): record bfloat16 operator coverage extension (Riley)
justinchuby Jul 23, 2026
63c771b
fix(shape-inference): allow dynamic-rank If outputs
justinchuby Jul 23, 2026
3792f4c
docs(squad): record dynamic-rank If outputs shape-inference fix (Sand…
justinchuby Jul 23, 2026
bc680de
perf(cpu): streamline qwen3.5 recurrent decode
justinchuby Jul 23, 2026
88cba98
fix(cpu): guard qwen3.5 recurrent direct writes against aliased bindings
justinchuby Jul 23, 2026
8b8a742
docs(squad): record qwen3.5 recurrent streamline + aliasing guard (Do…
justinchuby Jul 23, 2026
78e4d0d
docs(squad): record clean load-gated native-vs-ORT scoreboard (Langston)
justinchuby Jul 23, 2026
0adb960
perf(cpu): vectorize 8-bit MatMulNBits decode GEMV (fp32-accurate)
justinchuby Jul 23, 2026
e3ad0e6
docs(squad): record qwen3-0.6b native-correct + 8-bit GEMV vectorizat…
justinchuby Jul 23, 2026
c74985a
docs(squad): record Phi-3.5 executor control-flow overhead diagnosis …
justinchuby Jul 23, 2026
615a1dc
docs(squad): record Phi-3.5 CPU KV host round-trip root cause (Tripp,…
justinchuby Jul 23, 2026
ca16c3b
feat(engine): derive per-layer KV/state metadata from graph for hybri…
justinchuby Jul 23, 2026
6528499
docs(squad): record hybrid SSM qwen3.5-2b native enablement (Cooper, …
justinchuby Jul 23, 2026
ad0315d
perf(executor): attribute Phi-3.5 decode overhead to CPU KV host roun…
justinchuby Jul 23, 2026
3dde516
perf(executor): zero-copy hand-off of produced host outputs (eliminat…
justinchuby Jul 23, 2026
7d9364e
docs(squad): record zero-copy output hand-off + phase profiler (Tripp…
justinchuby Jul 23, 2026
b8cdcbc
fix(session): validate If signatures before EP optimization
justinchuby Jul 23, 2026
2a3412d
docs(squad): record control-flow validation-ordering fix (Sidle, Gris…
justinchuby Jul 24, 2026
0281675
feat(ep-cpu): in-place append-only GQA path for persistent present==p…
justinchuby Jul 23, 2026
a5ac872
perf(engine): wire persistent in-place CPU KV cache (eliminate input-…
justinchuby Jul 24, 2026
53ecf1b
docs(squad): record in-place CPU KV cache (Stokes, Messer-reviewed 🟢,…
justinchuby Jul 24, 2026
b8f0bc4
Merge remote-tracking branch 'origin/main' into perf/merge-main
justinchuby Jul 24, 2026
5ec5274
docs(squad): record origin/main merge into PR #105 (Taylor, Duquesne-…
justinchuby Jul 24, 2026
0312a9b
perf(cpu-ep): extend bf16 support across remaining capable kernels
justinchuby Jul 24, 2026
6811fa5
docs(squad): record bf16 op coverage merge (Grissom, Sanders review)
justinchuby Jul 24, 2026
e0cfd66
test(cpu): lock Phi-3.5 int4 native decode as more-accurate than ORT
justinchuby Jul 24, 2026
8d4c4fd
docs(squad): record Phi-3.5 divergence verdict (native more accurate,…
justinchuby Jul 24, 2026
1a0f101
docs(squad): record Gemma-2-2B f32 native-slower gap + f32 GEMM opt l…
justinchuby Jul 24, 2026
21942a6
docs(squad): record vision CNN f32 gap + missing Resize kernel (Curtis)
justinchuby Jul 24, 2026
cf96162
feat(cpu): implement ONNX Resize kernel
justinchuby Jul 24, 2026
32d68a2
docs(squad): record Resize kernel merge (Hawkes, Bonasera review)
justinchuby Jul 24, 2026
9cd1b1b
perf(cpu): route dense f32 GEMM through multithreaded MLAS by default
justinchuby Jul 24, 2026
b3f9430
docs(squad): record f32 GEMM MLAS-default win (Hodges, Danville review)
justinchuby Jul 24, 2026
30cebb6
docs(squad): record native-vs-ORT scoreboard (4/5 wins, qwen3 8-bit a…
justinchuby Jul 24, 2026
cf04f7b
docs(squad): record filing ORT correctness issue #29849
justinchuby Jul 24, 2026
4b30a0e
perf(cpu): int16-activation fast path for 8-bit MatMulNBits decode
Jul 24, 2026
a41b20a
docs(squad): record int16-activation 8-bit fast path (Ross, Vecchio r…
justinchuby Jul 24, 2026
ca885d7
perf(cpu): parallelize MLAS NCHWc conv/pool/reorder (Ecklie)
Jul 24, 2026
fc8a72f
docs(squad): record MLAS conv/pool/reorder parallelization (Ecklie, K…
justinchuby Jul 24, 2026
cc4f3ab
merge: integrate origin/main (CUDA + parity track) into CPU-EP perf b…
justinchuby Jul 24, 2026
6b59a9b
docs(squad): record origin/main merge into PR #105 (Willows, Duquesne…
justinchuby Jul 24, 2026
0478190
Merge remote-tracking branch 'origin/main' into perf/cpu-ep-mlas
justinchuby Jul 24, 2026
e76dad2
docs(squad): record profile_native bench merge with origin/main d03261c
justinchuby Jul 24, 2026
4e73898
perf(cpu): native avx512_bf16 GEMM via _mm512_dpbf16_ps (Caine)
justinchuby Jul 24, 2026
8ad00ad
docs(squad): record native avx512_bf16 GEMM (Caine, Sorenson review)
justinchuby Jul 24, 2026
58d5d6e
perf(cpu): widen VNNI/int16 decode dots to true 512-bit on AVX-512 (D…
justinchuby Jul 24, 2026
0a6b1e3
docs(squad): record 512-bit VNNI/int16 widening (Delko, Flack review …
justinchuby Jul 24, 2026
37ee582
perf(cpu): speed up int4 decode unpack ~1.4x (deinterleave-once + per…
justinchuby Jul 24, 2026
0841d79
docs(squad): record int4 unpack ~1.45x win (Bishop, Ferro review 🟢, l…
justinchuby Jul 24, 2026
9c29cc3
fix(cpu): gate NCHWc-via-MLAS path behind `mlas` feature + workspace fmt
justinchuby Jul 24, 2026
cd07d49
fix(cpu): reconcile int4 formatting after no-mlas fmt cherry-pick + r…
justinchuby Jul 24, 2026
16a0fae
perf(cpu): make persistent SPMD decode auto-enable contention-aware
justinchuby Jul 24, 2026
d85817d
perf(cpu): fmt fixup for SPMD contention cherry-pick + record decision
justinchuby Jul 24, 2026
adcfc5f
fix(cpu): clear EP clippy warnings
justinchuby Jul 24, 2026
386be50
docs(squad): record ep-cpu clippy cleanup (Crowe, Spunkmeyer review 🟢…
justinchuby Jul 24, 2026
bd45197
test(engine): lock qwen3-0.6b native decode to fp32-oracle argmax at …
justinchuby Jul 24, 2026
fef95ed
docs(squad): record qwen3-0.6b divergence adjudication (Hudson, Vasqu…
justinchuby Jul 24, 2026
643c4c6
fix(executor): restore early rejection before EP passes and host copy
justinchuby Jul 24, 2026
c321385
style: rustfmt qwen3_0_6b_divergence test imports
justinchuby Jul 24, 2026
01ad8d8
docs(squad): record executor early-reject fix (Dietrich, Hicks review…
justinchuby Jul 24, 2026
897ec52
chore(squad): scribe log — divergence + executor batch (orchestration…
justinchuby Jul 24, 2026
0575c20
fix(mlas): make cross-CPU SQNBit/kernel tests portable + guard AVX2 M…
justinchuby Jul 24, 2026
e649a00
docs(squad): record mlas cross-CPU fix + AVX2 M=1 asym int8 guard (Ri…
justinchuby Jul 24, 2026
ed9f997
merge: integrate origin/main (bf16 RoPE/norms, glm-5.2, executor) int…
justinchuby Jul 24, 2026
a352686
perf(cpu): fix batched MatMulNBits prefill via cached Nk dequant + ML…
justinchuby Jul 24, 2026
f5c6753
docs(squad): record batched prefill fix (Burke, Gorman2 review 🟢)
justinchuby Jul 24, 2026
891fe32
docs(squad): merge decisions inbox + session/orchestration logs
justinchuby Jul 24, 2026
c3f982e
docs(squad): archive decisions.md entries older than 7 days
justinchuby Jul 24, 2026
c7f73e4
fix(cpu): gate x86-only helpers to green ARM cross-platform CI (Hasfo…
justinchuby Jul 24, 2026
ddb5ccb
perf(cpu): make persistent SPMD decode pool opt-in; default flat path…
justinchuby Jul 24, 2026
b30bb26
test(engine): lock int4 CPU decode token divergences vs ORT
justinchuby Jul 24, 2026
32685eb
perf(ep-cpu): multi-accumulator int4 M=1 GEMV decode kernel
justinchuby Jul 24, 2026
91dbc90
bench: add generic native-vs-ORT inference benchmark
justinchuby Jul 24, 2026
1b880c9
perf(cpu): numa-split decode sizing fix + non-vacuous partition test …
justinchuby Jul 24, 2026
a37237d
perf(matmul_nbits): zero-copy f32 activation borrow in M>1 prefill path
justinchuby Jul 24, 2026
921b8a9
Merge remote-tracking branch 'origin/main' into perf/cpu-ep-mlas
justinchuby Jul 24, 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
2 changes: 2 additions & 0 deletions .squad/agents/chew/history.md
Original file line number Diff line number Diff line change
Expand Up @@ -109,3 +109,5 @@ Recorded the integration rejection and subsequent approval after dependency meta

### 2026-07-22T14:59:36+0000 — WP-B landed
WP-B landed: Chew's rejection of loader-IR shape authority directly informed the final Sapper WP-B3 v3 fix.

2026-07-22T22:15:00Z — Approved Zhora’s `f8848c9` MatMulNBits f16/bf16 and topology tuning with non-blocking parity/regression follow-ups.
2 changes: 2 additions & 0 deletions .squad/agents/gaff/history.md
Original file line number Diff line number Diff line change
Expand Up @@ -133,3 +133,5 @@ CI now covers all 27 offline crates with warnings-as-errors and native Windows A

## 2026-07-22T21-35-00Z — ORT CUDA attention review
Rejected Howie's ORT CUDA attention branch `7ff33496bda2` because `ONNX_GENAI_CUDA_ATTENTION` bypassed the typed `RuntimeConfig` registry. Named Deckard as reviser and locked Howie out of this artifact.

2026-07-22T22:15:00Z — Approved Tyrell’s `122b31a` cross-OS affinity and safe auto-enable after N1/N2 fixes; Windows >64-CPU validation remains non-blocking.
2 changes: 2 additions & 0 deletions .squad/agents/holden/history.md
Original file line number Diff line number Diff line change
Expand Up @@ -128,3 +128,5 @@ CI now covers all 27 offline crates with warnings-as-errors and native Windows A

- Holden reviewed Sapper's native CUDA fp16 int4 GEMV explicit-zero-point fix and returned 🟢 GREEN across all five criteria: SM-portability, capture-safety, symmetric no-regress, genericity, and correctness.
- Validation evidence: 6/6 unit tests and 18/18 `matmul_nbits_gpu` integration tests passed; Sapper's fix merged to main as `48de993`.

2026-07-22T22:15:00Z — Delivered portable CPU GQA SIMD tests in `82e44be`; Roy approved scalar-fallback dispatch coverage on non-AVX2 x86 and ARM.
1 change: 1 addition & 0 deletions .squad/agents/pris/history.md
Original file line number Diff line number Diff line change
Expand Up @@ -123,6 +123,7 @@ CI now covers all 27 offline crates with warnings-as-errors and native Windows A

Pris authored WP-B1 optional-modality metadata schema support and Bryant approved it; the work landed on origin/main as `a71c6f3`. Rachael's WP-B design note remains active for WP-B2/WP-B3 follow-up reference.

- 2026-07-22T23:20:00Z — Authored the persistent SPMD decode-pool profile/implementation; the default-off lever advanced native M=1 int4 decode to about 17.3 tok/s with bit parity, then entered review and revision.
## 2026-07-24T15:10:00Z — Shared decode-lock helper revision

Revised the Phi decode lock after Holden's rejection: environment-gated real-model coverage and generalized `common/decode_lock.rs` across Phi and Qwen. Holden approved; Pris co-owns the shared helper with Batty and Leon.
2 changes: 2 additions & 0 deletions .squad/agents/roy/history.md
Original file line number Diff line number Diff line change
Expand Up @@ -128,5 +128,7 @@ Authored CUDA Gather, Shape, and Constant kernels; coverage reached 65.

## 2026-07-21 — Wave-2 and CI milestone
CI now covers all 27 offline crates with warnings-as-errors and native Windows ARM64. Capture-safe native fp16 CUDA decode wave 2 stacked GQA prep fusion, warp-shuffle RMSNorm, and specialized down-projection GEMV on wave 1, reaching 663–672 tok/s on H200 versus ORT GenAI at 657, with zero fallbacks. All CUDA EP kernel work must remain correct and fast across supported SM architectures, not only sm_90.

2026-07-22T22:15:00Z — Approved Holden’s `82e44be` portable GQA SIMD test revision, including scalar fallback coverage.
### 2026-07-23T11:40Z
Scribe note: Roy generic lm_head Transpose-fold + dense fp16 M==1 GEMV fast path was independently approved by Buster and merged to main as 0a2422d; Iran also recorded PROGRESS.md follow-up a933ffe.
2 changes: 2 additions & 0 deletions .squad/agents/sebastian/history.md
Original file line number Diff line number Diff line change
Expand Up @@ -83,3 +83,5 @@ CI now covers all 27 offline crates with warnings-as-errors and native Windows A

## 2026-07-21T11:15:00Z — Wave-3 long-context GQA
- Raised capture-safe fp16 GQA `MAX_SPLITS` 8→16; Holden approved and `3b972bf` merged. Independent H200 review measured about 647→693 tok/s at 1024 tokens (+7.1%), flat at 256, with identical tokens and zero fallbacks.

- 2026-07-22T23:20:00Z — Revised the rejected persistent SPMD pool under lockout; `cee3c20` added real 31-worker parity, precedence diagnostics, and panic-safe poisoning, then merged after approval.
2 changes: 2 additions & 0 deletions .squad/agents/tyrell/history.md
Original file line number Diff line number Diff line change
Expand Up @@ -31,3 +31,5 @@ CI now covers all 27 offline crates with warnings-as-errors and native Windows A

### 2026-07-22T14:59:36+0000 — WP-B landed
WP-B landed: Progress docs should now treat WP-B as fully landed after WP-B3 `3d84b9b` plus clippy `6f217a4`.

2026-07-22T22:15:00Z — Landed `122b31a` cross-OS decode affinity and safe multi-NUMA auto-enable; Gaff approved after N1/N2 fixes, with Windows multigroup validation remaining non-blocking.
2 changes: 2 additions & 0 deletions .squad/agents/zhora/history.md
Original file line number Diff line number Diff line change
Expand Up @@ -109,3 +109,5 @@ Added Rayon N-range partitioning for native CPU MatMulNBits.

### 2026-07-22T14:59:36+0000 — WP-B landed
Zhora approved Rutger's clippy cleanup; required Clippy and targeted tests passed with scoped API-preserving fixes.

2026-07-22T22:15:00Z — Landed `f8848c9`: f16/bf16 MatMulNBits widening/narrowing and topology-derived tuning; Chew approved with non-blocking coverage follow-ups.
Loading
Loading