Update TensorBoard training metrics

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 1062967.6168812835,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 75338.9012150208,
-      "required_pages_per_second_upper_estimate": 0.006530185012618575
+      "required_token_positions_per_second": 75541.45076049298,
+      "required_pages_per_second_upper_estimate": 0.006547741493862974
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 1206672.263377472,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9417.3626518776,
-      "required_pages_per_second_upper_estimate": 0.0037567267639530876
+      "required_token_positions_per_second": 9442.681345061623,
+      "required_pages_per_second_upper_estimate": 0.0037668267692123913
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 410042.7581134458,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 18834.7253037552,
-      "required_pages_per_second_upper_estimate": 0.004357082219179862
+      "required_token_positions_per_second": 18885.362690123246,
+      "required_pages_per_second_upper_estimate": 0.004368796287328547
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 2163486.5624068812,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28252.087955632804,
-      "required_pages_per_second_upper_estimate": 0.00172040392924814
+      "required_token_positions_per_second": 28328.044035184867,
+      "required_pages_per_second_upper_estimate": 0.0017250292559821097
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 1546694.2737633225,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28252.087955632804,
-      "required_pages_per_second_upper_estimate": 0.0021749239376095313
+      "required_token_positions_per_second": 28328.044035184867,
+      "required_pages_per_second_upper_estimate": 0.0021807712468733345
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 22,
-      "required_token_positions_per_second": 28252.087955632804,
-      "required_pages_per_second_upper_estimate": 0.3507963787530303
+      "required_token_positions_per_second": 28328.044035184867,
+      "required_pages_per_second_upper_estimate": 0.35173949905242147
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,33 +158,27 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.0886703564531075,
-    "producer_p95_seconds": 0.21952342199983832,
-    "gpu_update_seconds": 0.9786179358997655,
-    "producer_batches_per_second": 11.277726175927135,
-    "gpu_batches_per_second": 1.021849246080469,
-    "producer_headroom": 11.036585111928568,
+    "producer_batch_seconds": 0.06498833267976067,
+    "producer_p95_seconds": 0.21245356899999024,
+    "gpu_update_seconds": 0.9759939643435948,
+    "producer_batches_per_second": 15.387377376299886,
+    "gpu_batches_per_second": 1.0245965001151933,
+    "producer_headroom": 15.017987446345868,
     "sustainable": true,
-    "buffer_seconds": 60.67431202578546,
+    "buffer_seconds": 60.511625789302876,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 84,
+    "upload_pages_per_commit": 85,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 283537.4830730475,
-    "upload_bytes_per_second": 5839943.91249327,
+    "generation_bytes_per_second": 282455.2507749909,
+    "upload_bytes_per_second": 5839741.981799227,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 60.840543452678375,
+    "estimated_full_shards_per_hour": 60.60832159459395,
     "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": 1.0866875020001316,
-      "bytes": 5928612,
-      "pages": 35,
-      "shards": 2
-    },
     {
       "seconds": 1.1460560750001605,
       "bytes": 6695792,
@@ -562,6 +556,12 @@
       "bytes": 4570486,
       "pages": 37,
       "shards": 2
+    },
+    {
+      "seconds": 1.065639946002193,
+      "bytes": 5791258,
+      "pages": 29,
+      "shards": 2
     }
   ],
   "cache_key": {

metrics/void-byte-v1/1790519810984249790-cb9c9513/train.jsonl

@@ -158,3 +158,4 @@
 {"step": 130800, "loss": 0.614993249066174, "lr": 0.0003, "grad_norm": 0.13111472899018056, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 19142127551, "tokens_per_second": 148461.5193809926, "data_wait_s": 1.2720000086119398e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 130900, "loss": 0.627563763409853, "lr": 0.0003, "grad_norm": 0.13571884190445932, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 19156667235, "tokens_per_second": 148638.82289956714, "data_wait_s": 1.6165999113582075e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 131000, "loss": 0.6175642311573029, "lr": 0.0003, "grad_norm": 0.1527774523782554, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 19171133531, "tokens_per_second": 148219.90765500045, "data_wait_s": 1.1596999684115872e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 131100, "loss": 0.622124794870615, "lr": 0.0003, "grad_norm": 0.1508883228916356, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 19185718227, "tokens_per_second": 148411.04620728875, "data_wait_s": 1.0424999345559627e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790519810.cad0b19a-59f2-43f0-8fb4-2a94097843e2-tjh7c.1336.0.1790519810984249790-cb9c9513

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7b33c48302ff0037450979ef07deafe85c3efd48ab7dc622ace2a461fd17ac43
-size 341394
+oid sha256:f383c73792659cd2b5726125493dd8ee2ba9be5d1c3f3b7d17faa6de95b6b62d
+size 343512