Update TensorBoard training metrics

benchmark_cache/performance.json

@@ -49,11 +49,11 @@
   "selected": {
     "microbatch": 90,
     "mode": "compiled",
-    "microbatch_seconds": 0.944348663000028,
-    "peak_gb": 89.37790203094482,
+    "microbatch_seconds": 0.9447870520000379,
+    "peak_gb": 89.37766551971436,
     "fits_budget": true,
-    "estimated_update_seconds": 0.944348663000028,
-    "input_tokens_per_second": 195182.1474649502
+    "estimated_update_seconds": 0.9447870520000379,
+    "input_tokens_per_second": 195091.58133550777
   },
   "measurements": [
     {

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 598315.063954628,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 78072.85898598007,
-      "required_pages_per_second_upper_estimate": 0.005401261538641098
+      "required_token_positions_per_second": 78036.6325342031,
+      "required_pages_per_second_upper_estimate": 0.005398755308650219
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 8533.957213384612,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9759.107373247509,
-      "required_pages_per_second_upper_estimate": 0.004085934463842666
+      "required_token_positions_per_second": 9754.579066775388,
+      "required_pages_per_second_upper_estimate": 0.0040840385564845805
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 380144.35761751566,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19518.214746495018,
-      "required_pages_per_second_upper_estimate": 0.004065129401757958
+      "required_token_positions_per_second": 19509.158133550776,
+      "required_pages_per_second_upper_estimate": 0.004063243148120618
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 43742.16556775435,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29277.322119742526,
-      "required_pages_per_second_upper_estimate": 0.0017551307571961265
+      "required_token_positions_per_second": 29263.737200326166,
+      "required_pages_per_second_upper_estimate": 0.0017543163620200866
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 22018.86276523721,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29277.322119742526,
-      "required_pages_per_second_upper_estimate": 0.0022761830174795844
+      "required_token_positions_per_second": 29263.737200326166,
+      "required_pages_per_second_upper_estimate": 0.002275126849748708
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 910315.184299418,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 23,
-      "required_token_positions_per_second": 29277.322119742526,
-      "required_pages_per_second_upper_estimate": 0.37553805261274903
+      "required_token_positions_per_second": 29263.737200326166,
+      "required_pages_per_second_upper_estimate": 0.375363799852826
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,15 +158,15 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 1.3731632000000218,
-    "producer_p95_seconds": 2.731159453000032,
-    "gpu_update_seconds": 0.944348663000028,
-    "producer_batches_per_second": 0.72824555741079,
-    "gpu_batches_per_second": 1.0589309215763356,
-    "producer_headroom": 0.6877177184765897,
-    "sustainable": false,
-    "buffer_seconds": 60.43831443200179,
-    "buffer_drain_seconds": 193.53744354999856,
+    "producer_batch_seconds": 0.5398314225000149,
+    "producer_p95_seconds": 1.0657202160000452,
+    "gpu_update_seconds": 0.9447870520000379,
+    "producer_batches_per_second": 1.852430144523446,
+    "gpu_batches_per_second": 1.0584395688775379,
+    "producer_headroom": 1.7501520152803105,
+    "sustainable": true,
+    "buffer_seconds": 60.466371328002424,
+    "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 7,
     "max_pending_pages": 1681,
@@ -221,7 +221,7 @@
   },
   "cache_reused": true,
   "resume_probe_seconds": [
-    2.731159453000032,
-    0.015166947000011533
+    1.0657202160000452,
+    0.013942628999984663
   ]
 }

runs/void-byte-v1/events.out.tfevents.1790956870.abdf5fd7-18c6-4e42-b51e-c994b21d5ba1-dh6p5.3284.0.1790956870018829698-ff939fa5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a488995a361c029ba8a0dc405a1966498ee3ae4622824908dd932508c06acf51
+size 165