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": 77287.41923569981,
-      "required_pages_per_second_upper_estimate": 0.005346922994245402
+      "required_token_positions_per_second": 77317.05426615606,
+      "required_pages_per_second_upper_estimate": 0.005348973214415109
     },
     {
       "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": 9660.927404462476,
-      "required_pages_per_second_upper_estimate": 0.004044828561143261
+      "required_token_positions_per_second": 9664.631783269508,
+      "required_pages_per_second_upper_estimate": 0.004046379507193538
     },
     {
       "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": 19321.854808924953,
-      "required_pages_per_second_upper_estimate": 0.004024232805121897
+      "required_token_positions_per_second": 19329.263566539015,
+      "required_pages_per_second_upper_estimate": 0.004025775853950829
     },
     {
       "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": 28982.78221338743,
-      "required_pages_per_second_upper_estimate": 0.0017374735395465346
+      "required_token_positions_per_second": 28993.89534980852,
+      "required_pages_per_second_upper_estimate": 0.0017381397551062025
     },
     {
       "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": 28982.78221338743,
-      "required_pages_per_second_upper_estimate": 0.0022532838353046063
+      "required_token_positions_per_second": 28993.89534980852,
+      "required_pages_per_second_upper_estimate": 0.0022541478327798253
     },
     {
       "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": 28982.78221338743,
-      "required_pages_per_second_upper_estimate": 0.37176001094633765
+      "required_token_positions_per_second": 28993.89534980852,
+      "required_pages_per_second_upper_estimate": 0.3719025583279912
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.04878984994532232,
-    "producer_p95_seconds": 0.13665231499999209,
-    "gpu_update_seconds": 0.9539456839043258,
-    "producer_batches_per_second": 20.496066315446296,
-    "gpu_batches_per_second": 1.0482777131578207,
-    "producer_headroom": 19.55213399863683,
+    "producer_batch_seconds": 0.04772446601562841,
+    "producer_p95_seconds": 0.12866096200014,
+    "gpu_update_seconds": 0.9535800438826717,
+    "producer_batches_per_second": 20.95361317762106,
+    "gpu_batches_per_second": 1.0486796639832363,
+    "producer_headroom": 19.98094737341642,
     "sustainable": true,
-    "buffer_seconds": 60.09857808597253,
+    "buffer_seconds": 60.075542764608315,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 64,
+    "upload_pages_per_commit": 67,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791143623306220418-eb05a537/train.jsonl

@@ -9,3 +9,4 @@
 {"step": 685900, "loss": 0.5103606581687927, "lr": 8e-05, "grad_norm": 0.20037664165679023, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100397882630, "tokens_per_second": 153820.61905108768, "data_wait_s": 1.2389000175971887e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 686000, "loss": 0.5892570614814758, "lr": 8e-05, "grad_norm": 0.21885647328773225, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100412502302, "tokens_per_second": 152886.282932156, "data_wait_s": 1.1249999943174771e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 686100, "loss": 0.5920264720916748, "lr": 8e-05, "grad_norm": 0.2095812736589121, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100427191750, "tokens_per_second": 153705.14991769212, "data_wait_s": 1.0631999884935794e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 686200, "loss": 0.5173473358154297, "lr": 8e-05, "grad_norm": 0.21569271361113784, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100441963613, "tokens_per_second": 155056.08306677538, "data_wait_s": 1.6467000023112632e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791143623.3eb6ddb3-ffce-42e5-bd48-c0d653189e98-8pb7d.369.0.1791143623306220418-eb05a537

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d1a3f4d970d7d7420027c704cb85123d170a77a7a91df41dff5f3b2df46ac7de
-size 23463
+oid sha256:2c6105cd8efa73dacf38995acd57bad5eaadf923d49da0e828f8d8b02d90f637
+size 25581