PLI Labs: standalone docs, website links, newest results
Browse filesDocs restructured (status, when to use, deploy, rollback, limits); links to the standalone GitHub repo and proprietarylegal.ai.
README.md
CHANGED
|
@@ -178,11 +178,11 @@ Changing a shell variable alone does not reconfigure the running engine. Keep th
|
|
| 178 |
|
| 179 |
- The combined package is an overview and configuration index. Implementation trees live in the standalone repositories; the historical hub folders are snapshots.
|
| 180 |
- The first long request after restart can still pay lazy-loading cost despite a short warm-up. Production counters cover a bounded check, not extended availability.
|
| 181 |
-
-
|
| 182 |
-
- Prefix storage:
|
| 183 |
-
- Prefill ordering: independent
|
| 184 |
- Native drafter fidelity: online-block attention and FP32 head corrections are implemented in a development candidate, but the strict stage-zero attention gate still fails. Historical speed figures predate that correction and are not a qualified gain for the corrected worker.
|
| 185 |
-
- Drafter self-distillation
|
| 186 |
- Further upstream adoption is offline, opt-in work: target MoE row splitting, replay/upload changes, Engram lookahead, and separate configuration candidates. It has no GB10 performance qualification and should be tested independently before combination.
|
| 187 |
- One x86 host is a shared performance dependency. Recovery and target-only behavior must be tested under your own operational constraints.
|
| 188 |
|
|
|
|
| 178 |
|
| 179 |
- The combined package is an overview and configuration index. Implementation trees live in the standalone repositories; the historical hub folders are snapshots.
|
| 180 |
- The first long request after restart can still pay lazy-loading cost despite a short warm-up. Production counters cover a bounded check, not extended availability.
|
| 181 |
+
- KV packing + 1M: GB10 validation in progress.
|
| 182 |
+
- Prefix storage: cleared for gated GB10 testing only.
|
| 183 |
+
- Prefill ordering: an independent check found a scheduler liveness bug; it is fixed (45 CPU tests) and under a second independent check.
|
| 184 |
- Native drafter fidelity: online-block attention and FP32 head corrections are implemented in a development candidate, but the strict stage-zero attention gate still fails. Historical speed figures predate that correction and are not a qualified gain for the corrected worker.
|
| 185 |
+
- Drafter self-distillation: a fast training path is built, but it does not yet match the exact reference math closely enough, so it is not used for training yet.
|
| 186 |
- Further upstream adoption is offline, opt-in work: target MoE row splitting, replay/upload changes, Engram lookahead, and separate configuration candidates. It has no GB10 performance qualification and should be tested independently before combination.
|
| 187 |
- One x86 host is a shared performance dependency. Recovery and target-only behavior must be tested under your own operational constraints.
|
| 188 |
|