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": 77167.53087859084,
-      "required_pages_per_second_upper_estimate": 0.005338628839521289
+      "required_token_positions_per_second": 77255.09039067777,
+      "required_pages_per_second_upper_estimate": 0.005344686409733521
     },
     {
       "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": 9645.941359823855,
-      "required_pages_per_second_upper_estimate": 0.004038554217197268
+      "required_token_positions_per_second": 9656.886298834721,
+      "required_pages_per_second_upper_estimate": 0.004043136634604801
     },
     {
       "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": 19291.88271964771,
-      "required_pages_per_second_upper_estimate": 0.0040179904093425945
+      "required_token_positions_per_second": 19313.772597669442,
+      "required_pages_per_second_upper_estimate": 0.004022549493659618
     },
     {
       "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": 28937.824079471568,
-      "required_pages_per_second_upper_estimate": 0.0017347783680653746
+      "required_token_positions_per_second": 28970.658896504163,
+      "required_pages_per_second_upper_estimate": 0.001736746765210616
     },
     {
       "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": 28937.824079471568,
-      "required_pages_per_second_upper_estimate": 0.0022497885381425804
+      "required_token_positions_per_second": 28970.658896504163,
+      "required_pages_per_second_upper_estimate": 0.002252341300741768
     },
     {
       "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": 28937.824079471568,
-      "required_pages_per_second_upper_estimate": 0.3711833362767482
+      "required_token_positions_per_second": 28970.658896504163,
+      "required_pages_per_second_upper_estimate": 0.3716045060543626
     }
   ],
   "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": 704566,
-    "producer_batch_seconds": 0.05561574154687143,
-    "producer_p95_seconds": 0.14380779100019936,
-    "gpu_update_seconds": 0.9554277448114502,
-    "producer_batches_per_second": 17.980520841517993,
-    "gpu_batches_per_second": 1.046651623244776,
-    "producer_headroom": 17.179088478146816,
+    "producer_batch_seconds": 0.04962241045311622,
+    "producer_p95_seconds": 0.1460787609998988,
+    "gpu_update_seconds": 0.954344880410581,
+    "producer_batches_per_second": 20.152185088727414,
+    "gpu_batches_per_second": 1.0478392251339759,
+    "producer_headroom": 19.232134668513456,
     "sustainable": true,
-    "buffer_seconds": 60.19194792312136,
+    "buffer_seconds": 60.1237274658666,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 44,
+    "upload_pages_per_commit": 52,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790956870018829698-ff939fa5/train.jsonl

@@ -3,3 +3,4 @@
 {"step": 525300, "loss": 0.5749369859695435, "lr": 0.0003, "grad_norm": 0.14512793442678665, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 76886579996, "tokens_per_second": 153859.12318553214, "data_wait_s": 9.031999979924876e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 525400, "loss": 0.5549130439758301, "lr": 0.0003, "grad_norm": 0.18048690312730425, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 76901085601, "tokens_per_second": 152199.06208305917, "data_wait_s": 1.6654000091875787e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 525500, "loss": 0.5804863572120667, "lr": 0.0003, "grad_norm": 0.17124735976777736, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 76915582593, "tokens_per_second": 152172.4227502477, "data_wait_s": 1.25609999486187e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 525600, "loss": 0.5729817748069763, "lr": 0.0003, "grad_norm": 0.17505624514239204, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 76930290681, "tokens_per_second": 154543.33481110237, "data_wait_s": 1.9686999621626455e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1790956870.abdf5fd7-18c6-4e42-b51e-c994b21d5ba1-dh6p5.3284.0.1790956870018829698-ff939fa5

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54504d17ff63d231685119920fb437bb68f51ecaaa59112e9e933cd2b8d84a75
-size 10755
+oid sha256:0f0a06081f16710421f7cdc9b184d0b624bd0333461a683c94babe0300a6ac30
+size 12873