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": 77514.11911580345,
-      "required_pages_per_second_upper_estimate": 0.005362606617967164
+      "required_token_positions_per_second": 77515.96942355663,
+      "required_pages_per_second_upper_estimate": 0.005362734626550835
     },
     {
       "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": 9689.264889475431,
-      "required_pages_per_second_upper_estimate": 0.004056692874364207
+      "required_token_positions_per_second": 9689.496177944578,
+      "required_pages_per_second_upper_estimate": 0.004056789710016387
     },
     {
       "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": 19378.529778950862,
-      "required_pages_per_second_upper_estimate": 0.004036036706758826
+      "required_token_positions_per_second": 19378.992355889157,
+      "required_pages_per_second_upper_estimate": 0.004036133049336098
     },
     {
       "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": 29067.794668426293,
-      "required_pages_per_second_upper_estimate": 0.001742569906419612
+      "required_token_positions_per_second": 29068.488533833737,
+      "required_pages_per_second_upper_estimate": 0.001742611502591317
     },
     {
       "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": 29067.794668426293,
-      "required_pages_per_second_upper_estimate": 0.0022598931797536068
+      "required_token_positions_per_second": 29068.488533833737,
+      "required_pages_per_second_upper_estimate": 0.002259947124737044
     },
     {
       "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": 29067.794668426293,
-      "required_pages_per_second_upper_estimate": 0.37285045944031364
+      "required_token_positions_per_second": 29068.488533833737,
+      "required_pages_per_second_upper_estimate": 0.37285935960074573
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,14 +158,14 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.055720334757825896,
-    "producer_p95_seconds": 0.17472467800007507,
-    "gpu_update_seconds": 0.9511557486688701,
-    "producer_batches_per_second": 17.946769421724454,
-    "gpu_batches_per_second": 1.0513525270698167,
-    "producer_headroom": 17.070172905507906,
+    "producer_batch_seconds": 0.05134338964063545,
+    "producer_p95_seconds": 0.13241987099991093,
+    "gpu_update_seconds": 0.9511330445619701,
+    "producer_batches_per_second": 19.476703953502817,
+    "gpu_batches_per_second": 1.051377623474889,
+    "producer_headroom": 18.524936729327294,
     "sustainable": true,
-    "buffer_seconds": 60.873967914807686,
+    "buffer_seconds": 60.87251485196609,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 79,

metrics/void-byte-v1/1791254470898546162-7724d9f6/train.jsonl

@@ -25,3 +25,4 @@
 {"step": 762500, "loss": 0.5541641116142273, "lr": 3e-05, "grad_norm": 0.2464639528453497, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111607380573, "tokens_per_second": 152839.79255499403, "data_wait_s": 1.392800004396122e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 762600, "loss": 0.5683630108833313, "lr": 3e-05, "grad_norm": 0.2581159766337038, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111621971646, "tokens_per_second": 152872.30040918954, "data_wait_s": 9.875000159809133e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 762700, "loss": 0.5494645833969116, "lr": 3e-05, "grad_norm": 0.2529401843485463, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111636742854, "tokens_per_second": 154850.33984594542, "data_wait_s": 1.6755000160628697e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 762800, "loss": 0.5381356477737427, "lr": 3e-05, "grad_norm": 0.2355803750415523, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111651462115, "tokens_per_second": 154221.52309780213, "data_wait_s": 2.231799999208306e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791254470.1e0c9aa3-e42e-441e-9e17-2e25895f438d-9bhv5.366.0.1791254470898546162-7724d9f6

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ba2a6638997ca569282ce13a3bec7a51691208f86c7d0e6b0e654ecba204184f
-size 57351
+oid sha256:cf281c424854f2137dc98d7ad4c4f4a2caebea3a0b724323edd84637a0df8c80
+size 59469