Joseph Magly
afe6f6465d
Merge current main for CLI layer-selection validation
2026-09-07 17:17:06 -04:00
Joseph Magly
92f17a2ad7
Merge pull request #200 from rajamukerji/fix/bayesian-silent-skip-and-warmstart-bounds
...
fix(bayesian): surface skipped optimization and clamp warm start to its search space
2026-09-07 17:16:37 -04:00
Joseph Magly
b54bfe3955
style(test): remove trailing blank line from Optuna contracts
2026-09-07 16:26:56 -04:00
Joseph Magly
707e762904
fix(packaging): install and validate the Bayesian optimizer by default
2026-09-07 16:23:48 -04:00
Joseph Magly
40236a150e
fix(bayesian): reset attempt state and persist every skipped outcome
2026-09-07 16:21:29 -04:00
Joseph Magly
bbf06fd6c8
test(cli): verify layer selection reaches local and remote execution
2026-09-07 16:20:38 -04:00
Raja Mukerji
5fe46d2dca
feat(cli): expose --layer-selection so layer strategy is reachable without the Python API
2026-09-07 09:43:39 -07:00
Raja Mukerji
2d31d64857
fix(bayesian): surface skipped optimization and clamp warm start to search space
2026-09-07 00:36:39 -07:00
Joseph Magly
7c332b0e4e
Merge pull request #199 from jmagly/feat/distributed-checkpoint-preflight-public
...
feat: add safe distributed checkpoint intake and preflight
2026-09-04 19:53:42 -04:00
Joseph Magly
985c9e9363
feat: add safe distributed checkpoint intake and preflight
2026-09-04 19:43:53 -04:00
Joseph Magly
5cc43c6e52
Merge pull request #196 from younger-plinius/feat/115-offloaded-fused-experts
...
feat(abliteration): update offloaded fused MoE expert tensors in place (#115 )
2026-09-02 16:42:19 -04:00
Joseph Magly
131a4a8b55
docs: align fused offload claims with test contract
2026-09-02 14:32:29 -04:00
Pliny the Younger and Claude Fable 5.1
89b63ce216
feat(abliteration): update offloaded fused MoE expert tensors in place ( #115 )
...
transformers 5 stores routed experts as fused 3D parameters
(experts.gate_up_proj / experts.down_proj) for Mixtral, Qwen3-MoE,
DeepSeek-V3, GLM-4 MoE, Llama 4, gpt-oss and OLMoE. When such a layer is
CPU- or disk-offloaded by Accelerate, EXCISE failed closed with
"offloaded fused expert tensors are not yet a supported surgery layout",
so every frontier MoE checkpoint large enough to need offload could not be
abliterated at all.
Route the three fused projections and the fused bias projection through the
existing LogicalParameterTransaction adapter: a meta-resident fused
parameter is resolved to a private copy of its authoritative backing value,
projected per expert, and committed atomically; the live parameter stays on
meta and quantized or unknown backing layouts still fail before mutation.
Thread offload_roots through the MoE dispatch call sites so parent-prefixed
Accelerate hooks resolve, and drop the preflight rejection of 3D meta
parameters (validate_offloaded_parameters already resolves them).
Tests: fused 3D, bias, granular and selective-inversion projections on
offloaded backing stores, bounded norm restoration, parent-prefixed hook
resolution, quantized fail-closed, commit-failure rollback, and a full
offline pipeline run on a disk-offloaded tiny Mixtral fixture that verifies
the saved checkpoint carries the updated experts with no meta tensors.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com >
2026-09-02 11:54:14 -04:00
Joseph Magly
e39f908832
Merge pull request #192 from elder-plinius/docs/191-qwen38-e03-bnb4-release
...
docs: publish Qwen3.8 E03 4-bit release evidence
2026-08-30 13:21:41 -04:00
Joseph Magly
6755bd3df2
docs: record Qwen3.8 E03 4-bit release
2026-08-29 23:12:45 -04:00
Joseph Magly
cd8b0b785c
Merge pull request #189 from elder-plinius/feat/185-qwen38-experiment-protocol
...
feat: add held-out Qwen3.8 experiment protocol
2026-08-29 20:54:57 -04:00
Joseph Magly
db731bc350
docs: link published Qwen3.8 E03 model
2026-08-29 20:52:17 -04:00
Joseph Magly
dd1336f9de
docs: record Qwen3.8 qualification evidence
2026-08-29 19:33:24 -04:00
Joseph Magly
d6d4f5ba22
feat: evaluate saved Qwen3.8 checkpoints
2026-08-29 19:14:04 -04:00
Joseph Magly
95c0cd606d
fix(eval): score full anchored completions
2026-08-29 19:00:51 -04:00
Joseph Magly
4b737cba64
feat(qwen38): add E03 coherence recovery candidate
2026-08-29 18:45:57 -04:00
Joseph Magly
941869d909
feat: add tune-only Qwen3.8 E02 candidate ( #185 )
2026-08-29 03:54:53 -04:00
Joseph Magly
312e63dd00
fix: override broker loading deadline after base env ( #185 )
2026-08-29 01:38:32 -04:00
Joseph Magly
d3533d382f
fix: align broker pending window with cold loads ( #185 )
2026-08-29 01:36:37 -04:00
Joseph Magly
02cafde149
fix: make supervised pipeline timeout configurable ( #185 )
2026-08-29 01:12:34 -04:00
Joseph Magly
9e149bf656
feat: add held-out Qwen3.8 experiment protocol ( #185 )
2026-08-29 00:17:41 -04:00
Joseph Magly
fb38a3b0c8
Merge pull request #188 from jmagly/feat/184-durable-runs
...
feat: add durable reconnect-safe experiment runs (#184 )
2026-08-28 23:55:36 -04:00
Joseph Magly
e8a36d796c
feat: add durable reconnect-safe experiment runs ( #184 )
2026-08-28 23:50:45 -04:00
Joseph Magly
b36a5f7d3d
Merge pull request #187 from jmagly/fix/183-spectral-inconclusive
...
fix: make spectral evidence fail inconclusive (#183 )
2026-08-28 23:31:49 -04:00
Joseph Magly
41654bdcdf
fix: make spectral evidence fail inconclusive ( #183 )
2026-08-28 23:25:14 -04:00
Joseph Magly
ef82652f44
Merge pull request #186 from jmagly/fix/182-measured-kl-exact-rollback
...
fix: make KL optimization measured and reversible
2026-08-28 23:20:46 -04:00
Joseph Magly
1f428f60cf
fix: make KL optimization measured and reversible ( #182 )
2026-08-28 23:11:26 -04:00
Joseph Magly
173fbeff90
Merge pull request #181 from jmagly/fix/180-bounded-gpu-admission-errors
...
fix: surface bounded GPU admission failures
2026-08-28 10:48:48 -04:00
Joseph Magly
089c6eb5c0
docs: clarify shared-host admission budget
2026-08-28 10:46:24 -04:00
Joseph Magly
683b39ba53
fix: surface bounded GPU admission failures
2026-08-28 10:42:41 -04:00
Joseph Magly
a920391ae2
Merge pull request #179 from jmagly/fix/168-qwen38-reload-contract
...
fix(qwen): preserve validated runtime on checkpoint reload
2026-08-27 20:25:57 -04:00
Joseph Magly
691a5da746
fix(qwen): recognize saved text derivatives
2026-08-27 20:17:09 -04:00
Joseph Magly
44f7842c9c
fix(qwen): preserve validated runtime on checkpoint reload
2026-08-27 20:12:43 -04:00
Joseph Magly
fff7d4a24b
Merge pull request #178 from jmagly/fix/177-qwen-fast-runtime
...
fix: validate Qwen hybrid kernels and single-GPU placement
2026-08-27 01:14:17 -04:00
Joseph Magly
827ba8e106
fix: validate Qwen hybrid runtime placement
2026-08-27 01:12:20 -04:00
Joseph Magly
477b14fc61
Merge pull request #176 from jmagly/fix/175-qwen-multi-eos-probes
...
fix(qwen): preserve checkpoint multi-EOS stops
2026-08-26 23:47:09 -04:00
Joseph Magly
270db39e27
fix(qwen): preserve checkpoint multi-EOS stops
2026-08-26 23:45:01 -04:00
Joseph Magly
a1de7ff473
Merge pull request #174 from jmagly/fix/173-bounded-cuda-context-residue
...
fix(gpu): permit bounded CUDA context residue
2026-08-26 23:23:19 -04:00
Joseph Magly
2145a22d41
fix(gpu): permit bounded CUDA context residue
2026-08-26 23:21:03 -04:00
Joseph Magly
55adb487e2
Merge pull request #172 from jmagly/fix/171-benchmark-gpu-lifecycle
...
fix(gpu): broker benchmark model loads
2026-08-26 23:00:01 -04:00
Joseph Magly
f31d182456
test(gpu): cover benchmark lifecycle orchestration
2026-08-26 22:57:37 -04:00
Joseph Magly
6df26d07e7
test(gpu): keep benchmark contracts CPU-only
2026-08-26 22:51:13 -04:00
Joseph Magly
9ab258359e
fix(gpu): admit benchmark loads through lifecycle broker
2026-08-26 22:48:59 -04:00
Joseph Magly
5fb64fd7e8
Merge pull request #170 from jmagly/feat/168-qwen38-projection-contract
...
feat(surgery): validate Qwen3.8 residual-writer projection contract
2026-08-26 22:37:04 -04:00
Joseph Magly
5aa11828a1
feat(surgery): add validated Qwen3.8 output-writer contract
2026-08-26 22:32:14 -04:00
Joseph Magly