uph.git/adapt/fit_theta.py, branch master Unnamed repository; edit this file 'description' to name the repository. Fix two bugs: PEFT cleanup model corruption and K=16 OOM 2026-04-10T19:50:22+00:00 YurenHao0426 Blackhao0426@gmail.com 2026-04-10T19:50:22+00:00 112c5d354f36d6ea6e8049cf1aeaebeb9944aa02 Bug 1: PEFTBaseline.cleanup() corrupted wrapper.model after LoRA unload, causing 'Qwen2Model has no attribute prepare_inputs_for_generation' for subsequent methods. Fix: save reference to original model before wrapping, restore it directly in cleanup() instead of relying on unload(). Bug 2: fit_theta OOM at K=16 due to large logit chunks (128 × 151936 vocab). Fix: reduce CHUNK_SIZE from 128 to 32 (~4x less memory per chunk). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> 
Bug 1: PEFTBaseline.cleanup() corrupted wrapper.model after LoRA unload,
causing 'Qwen2Model has no attribute prepare_inputs_for_generation' for
subsequent methods. Fix: save reference to original model before wrapping,
restore it directly in cleanup() instead of relying on unload().

Bug 2: fit_theta OOM at K=16 due to large logit chunks (128 × 151936 vocab).
Fix: reduce CHUNK_SIZE from 128 to 32 (~4x less memory per chunk).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Initial commit: UPH project codebase and experiment results 2026-04-03T20:12:34+00:00 YurenHao0426 Blackhao0426@gmail.com 2026-04-03T20:12:34+00:00 8fe28101366dd32562b8c5534d7fe359b252bdf3 Includes model code, evaluation scripts, configs, analysis outputs, and experiment results for the User Prior Head personalization method. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> 
Includes model code, evaluation scripts, configs, analysis outputs,
and experiment results for the User Prior Head personalization method.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>