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
9e149bf656
feat: add held-out Qwen3.8 experiment protocol ( #185 )
2026-08-29 00:17:41 -04:00
Joseph Magly
5aa11828a1
feat(surgery): add validated Qwen3.8 output-writer contract
2026-08-26 22:32:14 -04:00
Joseph Magly
b8b7a12493
fix(pipeline): tolerate missing cancellation state
2026-08-24 14:36:39 -04:00
Joseph Magly
ccba4f5d85
fix(surgery): guard unvalidated Qwen hybrid layouts
2026-08-24 13:54:31 -04:00
Joseph Magly
d8d0a76231
fix(pipeline): fail closed on invalid output
2026-08-24 13:42:13 -04:00
Joseph Magly
ad883cef0b
feat: harden local runtime and verification
2026-08-23 11:16:10 -04:00
Joseph Magly
bd878363f8
test: lock Qwen2.5 projection contract
2026-08-21 19:09:34 -04:00
Joseph Magly
cf2b5ae869
test: cover single-prompt flattened routing
2026-08-21 14:43:00 -04:00
Joseph Magly
f2446f68a7
fix: preserve per-prompt MoE router observations
2026-08-21 14:33:13 -04:00
Joseph Magly
c74538a2be
test: harden checkpoint transaction reliability
2026-08-16 00:28:15 -04:00
Joseph Magly
c642cb3297
test: harden GPU memory utilization contract
2026-08-15 02:47:28 -04:00
Joseph Magly
18c910b529
test: validate refusal token limits
2026-08-14 20:03:46 -04:00
Joseph Magly
66c47249df
ci: add strict test evidence gates
2026-08-14 11:44:12 -04:00
Joseph Magly
bc891794db
test: cover offloaded state materialization ( #27 )
2026-08-14 10:26:56 -04:00
Aaron Meese
e1a0fe59f0
fix: REBIRTH save crash for CPU-offloaded models
...
Replace the manual safetensors/dat file materialization in
_gather_state_dict with accelerate's get_state_dict_offloaded_model().
The old code only handled disk-offloaded weights but failed for the 398
CPU-offloaded meta tensors managed by accelerate's AlignDevicesHook.
Pre-move all GPU tensors to CPU before materialization to prevent CUDA
OOM when align_module_device restores non-hooked params to CUDA during
its __exit__ cleanup.
Update test to verify the safety net catches unmaterialized meta tensors
after the accelerate path (the old test checked for a missing offload
directory, which this codepath no longer uses).
2026-08-14 10:22:48 -04:00
Joseph Magly
548092d5ca
ci: establish green validation baseline ( #59 )
2026-08-14 09:55:49 -04:00
pliny
66ea4a6f86
Add files via upload
2026-03-05 00:50:44 -08:00
Joseph Magly