summaryrefslogtreecommitdiff
AgeCommit message (Expand)Author
2026-08-02Add tests/test_loss_reweight.py: batch determinism + weighting semantics (run...Void Agent
2026-08-02loss_reweight: CPU generators for randint/randperm (CUDA generators unsupport...Void Agent
2026-08-02Add causal experiments: synthetic frequency-matched pair (Gemini design) + lo...Void Agent
2026-07-31jlens_v3: chunk VJP probes (16) to fit K2200 4GB; proxy chunk fix; e2e verifi...Void Agent
2026-07-31Add faithful J-lens (jlens_v3): W_U-probed residual Jacobian per paper; both-...Void Agent
2026-07-30GPT-2 J-lens: use local Shakespeare + fix summaryVoid Agent
2026-07-30GPT-2 Small J-lens: frequency + dimensionality samplingVoid Agent
2026-07-30Add Pythia-70m test scriptVoid Agent
2026-07-30Fix: lr→learning_rate in configure_optimizersVoid Agent
2026-07-30Add dimensional starvation experiment (d_model=16,32,64,128 vs 384)Void Agent
2026-07-30Add controlled frequency ablation experiment (doubled-q)Void Agent
2026-07-29jlens_v2: fix import path for src/ -> root layoutVoid Agent
2026-07-29Add jlens_v2.py: simplified gradient approach for J-lensVoid Agent
2026-07-29jlens.py: fix hooks — don't detach activations, preserve gradient graphVoid Agent
2026-07-29jlens.py: fix load_model for dict-format nanoGPT checkpointsVoid Agent
2026-07-29Fix jlens.py import path: src/ -> project rootVoid Agent
2026-07-29Maxwell GPU fixes: disable bf16 SDPA, force fp32Void Agent
2026-07-29Restructure: nanoGPT at root, custom code in src/Void Agent
2026-07-29Initial project setup: J-lens implementation for nanoGPTVoid Agent