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": 75978.98425466158,
-      "required_pages_per_second_upper_estimate": 0.006585665788219006
+      "required_token_positions_per_second": 76011.28355353956,
+      "required_pages_per_second_upper_estimate": 0.006588465409583924
     },
     {
       "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": 9497.373031832698,
-      "required_pages_per_second_upper_estimate": 0.0037886441007789606
+      "required_token_positions_per_second": 9501.410444192445,
+      "required_pages_per_second_upper_estimate": 0.0037902546849339575
     },
     {
       "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": 18994.746063665396,
-      "required_pages_per_second_upper_estimate": 0.004394100205715899
+      "required_token_positions_per_second": 19002.82088838489,
+      "required_pages_per_second_upper_estimate": 0.004395968174302692
     },
     {
       "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": 28492.119095498096,
-      "required_pages_per_second_upper_estimate": 0.0017350205663198728
+      "required_token_positions_per_second": 28504.231332577336,
+      "required_pages_per_second_upper_estimate": 0.0017357581380100019
     },
     {
       "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": 28492.119095498096,
-      "required_pages_per_second_upper_estimate": 0.002193402199205083
+      "required_token_positions_per_second": 28504.231332577336,
+      "required_pages_per_second_upper_estimate": 0.0021943346327442615
     },
     {
       "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": 28492.119095498096,
-      "required_pages_per_second_upper_estimate": 0.353776762177609
+      "required_token_positions_per_second": 28504.231332577336,
+      "required_pages_per_second_upper_estimate": 0.3539271556250833
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,14 +158,14 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 690670,
-    "producer_batch_seconds": 0.05150864582812531,
-    "producer_p95_seconds": 0.13797506999981124,
-    "gpu_update_seconds": 0.9703735937411735,
-    "producer_batches_per_second": 19.414216466431917,
-    "gpu_batches_per_second": 1.0305309279332355,
-    "producer_headroom": 18.839043002200604,
+    "producer_batch_seconds": 0.04951623901564517,
+    "producer_p95_seconds": 0.12430535700013934,
+    "gpu_update_seconds": 0.9699612551348209,
+    "producer_batches_per_second": 20.195394882152492,
+    "gpu_batches_per_second": 1.0309690152118538,
+    "producer_headroom": 19.58875056783597,
     "sustainable": true,
-    "buffer_seconds": 60.16316281195276,
+    "buffer_seconds": 60.1375978183589,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 69,

metrics/void-byte-v1/1790696019647648547-3fdea0a8/train.jsonl

@@ -15,3 +15,4 @@
 {"step": 276500, "loss": 0.6285677775740623, "lr": 0.0003, "grad_norm": 0.1593182684767307, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40471107394, "tokens_per_second": 149054.53261513283, "data_wait_s": 1.1183999959030189e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 276600, "loss": 0.5801319926977158, "lr": 0.0003, "grad_norm": 0.14238914330006738, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40485662186, "tokens_per_second": 149788.11770255343, "data_wait_s": 1.7440000192436855e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 276700, "loss": 0.6008296497166157, "lr": 0.0003, "grad_norm": 0.15391543790461298, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40500433646, "tokens_per_second": 152339.59975473586, "data_wait_s": 1.401899999109446e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 276800, "loss": 0.6234837397933006, "lr": 0.0003, "grad_norm": 0.17697912234679625, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40515274660, "tokens_per_second": 153003.5590317808, "data_wait_s": 7.983000159583753e-06, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790696019.3e6d470a-8cc7-4d7e-a224-51a0f9a2ce55-85btc.22407.0.1790696019647648547-3fdea0a8

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7b8df9df37b6f7e17a212024bfa88602c55df8261352a1c1d1ae824dd96741f5
-size 36171
+oid sha256:d62c2561033d31c187c0031d0d6ae23d16fe01805d30801aeb5c0b07560de956
+size 38289