Update TensorBoard training metrics

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": 77155.50268032493,
-      "required_pages_per_second_upper_estimate": 0.005337796700855988
+      "required_token_positions_per_second": 76375.6706640414,
+      "required_pages_per_second_upper_estimate": 0.005283846112509921
     },
     {
       "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": 9644.437835040617,
-      "required_pages_per_second_upper_estimate": 0.004037924722767694
+      "required_token_positions_per_second": 9546.958833005176,
+      "required_pages_per_second_upper_estimate": 0.003997112300208492
     },
     {
       "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": 19288.875670081234,
-      "required_pages_per_second_upper_estimate": 0.004017364120219128
+      "required_token_positions_per_second": 19093.91766601035,
+      "required_pages_per_second_upper_estimate": 0.003976759509359474
     },
     {
       "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": 28933.31350512185,
-      "required_pages_per_second_upper_estimate": 0.0017345079660203572
+      "required_token_positions_per_second": 28640.876499015525,
+      "required_pages_per_second_upper_estimate": 0.0017169768140297367
     },
     {
       "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": 28933.31350512185,
-      "required_pages_per_second_upper_estimate": 0.0022494378608268095
+      "required_token_positions_per_second": 28640.876499015525,
+      "required_pages_per_second_upper_estimate": 0.002226702169896493
     },
     {
       "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": 28933.31350512185,
-      "required_pages_per_second_upper_estimate": 0.3711254794720674
+      "required_token_positions_per_second": 28640.876499015525,
+      "required_pages_per_second_upper_estimate": 0.36737441155212897
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,33 +158,27 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.12189683281286534,
-    "producer_p95_seconds": 0.3881533629974001,
-    "gpu_update_seconds": 0.9555766917296106,
-    "producer_batches_per_second": 8.203658593289203,
-    "gpu_batches_per_second": 1.0464884803646501,
-    "producer_headroom": 7.839224938654487,
+    "producer_batch_seconds": 0.15020428112444506,
+    "producer_p95_seconds": 0.3366278989997227,
+    "gpu_update_seconds": 0.9653335854072185,
+    "producer_batches_per_second": 6.657599853438895,
+    "gpu_batches_per_second": 1.0359113317062907,
+    "producer_headroom": 6.426804736726741,
     "sustainable": true,
-    "buffer_seconds": 60.201331578965465,
+    "buffer_seconds": 60.81601588065477,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 83,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 274206.6202150226,
-    "upload_bytes_per_second": 4471118.07163161,
+    "generation_bytes_per_second": 273762.60107156524,
+    "upload_bytes_per_second": 4419757.457076919,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 58.838357494716725,
+    "estimated_full_shards_per_hour": 58.74308132276743,
     "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
   },
   "upload_measurements": [
-    {
-      "seconds": 2.0095185120007955,
-      "bytes": 12961722,
-      "pages": 36,
-      "shards": 4
-    },
     {
       "seconds": 1.4560003849983332,
       "bytes": 6473170,
@@ -562,6 +556,12 @@
       "bytes": 6406874,
       "pages": 38,
       "shards": 2
+    },
+    {
+      "seconds": 1.7032136210036697,
+      "bytes": 6726184,
+      "pages": 31,
+      "shards": 3
     }
   ],
   "cache_key": {