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": 75648.12362092586,
-      "required_pages_per_second_upper_estimate": 0.006556987627045403
+      "required_token_positions_per_second": 75798.6960801964,
+      "required_pages_per_second_upper_estimate": 0.006570038866192564
     },
     {
       "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": 9456.015452615733,
-      "required_pages_per_second_upper_estimate": 0.0037721459440784
+      "required_token_positions_per_second": 9474.83701002455,
+      "required_pages_per_second_upper_estimate": 0.003779654144736138
     },
     {
       "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": 18912.030905231466,
-      "required_pages_per_second_upper_estimate": 0.004374965509549278
+      "required_token_positions_per_second": 18949.6740200491,
+      "required_pages_per_second_upper_estimate": 0.004383673581666144
     },
     {
       "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": 28368.0463578472,
-      "required_pages_per_second_upper_estimate": 0.001727465187556279
+      "required_token_positions_per_second": 28424.511030073652,
+      "required_pages_per_second_upper_estimate": 0.00173090358984763
     },
     {
       "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": 28368.0463578472,
-      "required_pages_per_second_upper_estimate": 0.0021838507364053974
+      "required_token_positions_per_second": 28424.511030073652,
+      "required_pages_per_second_upper_estimate": 0.0021881975431776064
     },
     {
       "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": 28368.0463578472,
-      "required_pages_per_second_upper_estimate": 0.35223619402072587
+      "required_token_positions_per_second": 28424.511030073652,
+      "required_pages_per_second_upper_estimate": 0.3529372962746769
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,24 +158,24 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.10788504744539296,
-    "producer_p95_seconds": 0.35009541399995214,
-    "gpu_update_seconds": 0.9746176966589728,
-    "producer_batches_per_second": 9.269125088962486,
-    "gpu_batches_per_second": 1.0260433433827836,
-    "producer_headroom": 9.033853344248515,
+    "producer_batch_seconds": 0.09109087118764592,
+    "producer_p95_seconds": 0.34746942700076033,
+    "gpu_update_seconds": 0.9726816398265536,
+    "producer_batches_per_second": 10.978048480182103,
+    "gpu_batches_per_second": 1.0280856130669,
+    "producer_headroom": 10.678146197798933,
     "sustainable": true,
-    "buffer_seconds": 60.42629719285631,
+    "buffer_seconds": 60.306261669246325,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 371770.9009268763,
+    "generation_bytes_per_second": 372009.47732887085,
     "upload_bytes_per_second": 3680294.027406763,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 79.77338095526424,
+    "estimated_full_shards_per_hour": 79.82457389735788,
     "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": [

metrics/void-byte-v1/1790434090210070261-a05fc3fb/train.jsonl

@@ -321,3 +321,4 @@
 {"step": 72100, "loss": 0.605940455570817, "lr": 0.0003, "grad_norm": 0.14389794579566484, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 10552078089, "tokens_per_second": 149666.46635261903, "data_wait_s": 8.277002052636817e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 72200, "loss": 0.6321951113641262, "lr": 0.0003, "grad_norm": 0.1372418156646615, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 10566890635, "tokens_per_second": 152371.74784975115, "data_wait_s": 1.0890998964896426e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 72300, "loss": 0.6349485367536545, "lr": 0.0003, "grad_norm": 0.12504722232350515, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 10581358498, "tokens_per_second": 147930.35049789556, "data_wait_s": 8.243998308898881e-06, "gpu_peak_gb": 85.46427917480469}
+{"step": 72400, "loss": 0.6408884692937136, "lr": 0.0003, "grad_norm": 0.15654120584981696, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 10595843036, "tokens_per_second": 147119.27267972243, "data_wait_s": 1.3313001545611769e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790434090.372bd512-57cc-4928-b44a-3b375a0fa071-5spg5.451.0.1790434090210070261-a05fc3fb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c2dc85f0c23dc80d8a48736726f2e1d5775e7ddb6f659a8297ef2f9fb5b4ad4d
-size 688977
+oid sha256:72aebc25220ef663f53e0ab96f5195e0d06d4809f3d7a8776714634cd27471b7
+size 691095