<feed xmlns='http://www.w3.org/2005/Atom'>
<title>ept.git/ep_run/holo_ep.py, branch xiang</title>
<subtitle>Unnamed repository; edit this file 'description' to name the repository.
</subtitle>
<link rel='alternate' type='text/html' href='https://git.blackhao.com/ept.git/'/>
<entry>
<title>bp_lm twin control + env forensics probes + holo_a_track_avg (semi-convergence fix, ungated)</title>
<updated>2026-07-06T07:22:54+00:00</updated>
<author>
<name>Yuren Hao</name>
<email>yurenh2@illinois.edu</email>
</author>
<published>2026-07-06T07:22:54+00:00</published>
<link rel='alternate' type='text/html' href='https://git.blackhao.com/ept.git/commit/?id=8f43de671cf03ee20d5d652b6cd3ba575982a436'/>
<id>8f43de671cf03ee20d5d652b6cd3ba575982a436</id>
<content type='text'>
BP+EMA 2x2 (local, healthy env): qknorm {1.9951, 1.9977}, no-qknorm {1.9728,
1.9732} -&gt; parameterization-matched BP twin tops at ~1.97 vs EP 1.7888: the
equilibrium computation's iteration/depth dividend = 0.18 CE from identical
parameters. External anchor (tuned depth-1 BP, 1.7921): EP at parity.
bp_lm on 107/2.3.1 gave 1.9823 ~= local -&gt; plain backward exonerated on the
pascal env; the 107 divergence (pair/floss/resreg dead by step 600) narrows
to the EP reg/estimator loop. Single-step fingerprints all match (a/b/c/d) -&gt;
suspected intermittent kernel issue; discriminators in flight (torch-2.7 env
probe + delay_hr2's step-2000 reg-on transition).

Co-Authored-By: Claude Fable 5 &lt;noreply@anthropic.com&gt;
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
BP+EMA 2x2 (local, healthy env): qknorm {1.9951, 1.9977}, no-qknorm {1.9728,
1.9732} -&gt; parameterization-matched BP twin tops at ~1.97 vs EP 1.7888: the
equilibrium computation's iteration/depth dividend = 0.18 CE from identical
parameters. External anchor (tuned depth-1 BP, 1.7921): EP at parity.
bp_lm on 107/2.3.1 gave 1.9823 ~= local -&gt; plain backward exonerated on the
pascal env; the 107 divergence (pair/floss/resreg dead by step 600) narrows
to the EP reg/estimator loop. Single-step fingerprints all match (a/b/c/d) -&gt;
suspected intermittent kernel issue; discriminators in flight (torch-2.7 env
probe + delay_hr2's step-2000 reg-on transition).

Co-Authored-By: Claude Fable 5 &lt;noreply@anthropic.com&gt;
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
</pre>
</div>
</content>
</entry>
<entry>
<title>holofast: exact halved-jvp AEP track — 1.55x nudged phase, gradient-gate passed</title>
<updated>2026-07-05T08:53:57+00:00</updated>
<author>
<name>Yuren Hao</name>
<email>yurenh2@illinois.edu</email>
</author>
<published>2026-07-05T08:53:57+00:00</published>
<link rel='alternate' type='text/html' href='https://git.blackhao.com/ept.git/commit/?id=cbecb171b1af77fe6510fd60f1dfa4c7938a20d6'/>
<id>cbecb171b1af77fe6510fd60f1dfa4c7938a20d6</id>
<content type='text'>
holo_a_track computed the doubled-batch jvp/vjp on [v0; -v0] at a shared
anchor zbar — exact antisymmetric redundancy (phase deviations from the
common mode are exact negatives). holo_a_track_fast computes at batch B and
mirrors: single-eval parity 6e-7 (exact); trajectory-level 45% divergence
SHARED with the original's own FD noise floor (1e-6 state noise -&gt; 49%
self-divergence — the 2r=0.04 finite difference amplifies fp noise; training
averages it via pema/momentum). Ship gate: cos(EP,BPTT) orig vs fast
indistinguishable (0.907/0.912, 0.853/0.853, 0.918/0.918). Timing 6.43-&gt;4.16s
on the T2=40 nudged phase (contended GPU, relative). --holofast flag,
default off; queued ablation arms deliberately stay on orig for fidelity.

Co-Authored-By: Claude Fable 5 &lt;noreply@anthropic.com&gt;
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
holo_a_track computed the doubled-batch jvp/vjp on [v0; -v0] at a shared
anchor zbar — exact antisymmetric redundancy (phase deviations from the
common mode are exact negatives). holo_a_track_fast computes at batch B and
mirrors: single-eval parity 6e-7 (exact); trajectory-level 45% divergence
SHARED with the original's own FD noise floor (1e-6 state noise -&gt; 49%
self-divergence — the 2r=0.04 finite difference amplifies fp noise; training
averages it via pema/momentum). Ship gate: cos(EP,BPTT) orig vs fast
indistinguishable (0.907/0.912, 0.853/0.853, 0.918/0.918). Timing 6.43-&gt;4.16s
on the T2=40 nudged phase (contended GPU, relative). --holofast flag,
default off; queued ablation arms deliberately stay on orig for fidelity.

Co-Authored-By: Claude Fable 5 &lt;noreply@anthropic.com&gt;
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
</pre>
</div>
</content>
</entry>
<entry>
<title>Initial commit: ept — backprop-free equilibrium transformer (EP)</title>
<updated>2026-07-03T10:56:50+00:00</updated>
<author>
<name>Yuren Hao</name>
<email>yurenh2@illinois.edu</email>
</author>
<published>2026-07-03T10:56:50+00:00</published>
<link rel='alternate' type='text/html' href='https://git.blackhao.com/ept.git/commit/?id=b83947778e2c776f757a07d4719b7ce961d7ed55'/>
<id>b83947778e2c776f757a07d4719b7ce961d7ed55</id>
<content type='text'>
Code (ep_run/), organized docs (docs/{method,campaign,hardware,outreach,paper}),
analysis scripts (scripts/), ONBOARDING.md entry point. Large data/checkpoints
git-ignored (share separately).

Co-Authored-By: Claude Opus 4.8 &lt;noreply@anthropic.com&gt;
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
Code (ep_run/), organized docs (docs/{method,campaign,hardware,outreach,paper}),
analysis scripts (scripts/), ONBOARDING.md entry point. Large data/checkpoints
git-ignored (share separately).

Co-Authored-By: Claude Opus 4.8 &lt;noreply@anthropic.com&gt;
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
</pre>
</div>
</content>
</entry>
</feed>
