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": 75761.71551975409,
-      "required_pages_per_second_upper_estimate": 0.006566833484939798
+      "required_token_positions_per_second": 75648.93950886454,
+      "required_pages_per_second_upper_estimate": 0.006557058346144078
     },
     {
       "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": 9470.21443996926,
-      "required_pages_per_second_upper_estimate": 0.0037778101324275016
+      "required_token_positions_per_second": 9456.117438608067,
+      "required_pages_per_second_upper_estimate": 0.0037721866278155685
     },
     {
       "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": 18940.42887993852,
-      "required_pages_per_second_upper_estimate": 0.00438153488121047
+      "required_token_positions_per_second": 18912.234877216135,
+      "required_pages_per_second_upper_estimate": 0.0043750126948781465
     },
     {
       "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": 28410.643319907784,
-      "required_pages_per_second_upper_estimate": 0.0017300591190567806
+      "required_token_positions_per_second": 28368.352315824206,
+      "required_pages_per_second_upper_estimate": 0.0017274838187918411
     },
     {
       "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": 28410.643319907784,
-      "required_pages_per_second_upper_estimate": 0.002187129968460759
+      "required_token_positions_per_second": 28368.352315824206,
+      "required_pages_per_second_upper_estimate": 0.002183874289897413
     },
     {
       "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": 28410.643319907784,
-      "required_pages_per_second_upper_estimate": 0.3527651057266571
+      "required_token_positions_per_second": 28368.352315824206,
+      "required_pages_per_second_upper_estimate": 0.35223999299482484
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.05663594865622201,
-    "producer_p95_seconds": 0.16449844799990387,
-    "gpu_update_seconds": 0.9731564220028278,
-    "producer_batches_per_second": 17.656630174413795,
-    "gpu_batches_per_second": 1.0275840321147203,
-    "producer_headroom": 17.182663045159693,
+    "producer_batch_seconds": 0.061839020335966666,
+    "producer_p95_seconds": 0.17076587500014284,
+    "gpu_update_seconds": 0.974607185225122,
+    "producer_batches_per_second": 16.17101943994385,
+    "gpu_batches_per_second": 1.0260544095711879,
+    "producer_headroom": 15.760391738584405,
     "sustainable": true,
-    "buffer_seconds": 60.33569816417533,
+    "buffer_seconds": 60.42564548395757,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 84,
+    "upload_pages_per_commit": 82,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790519810984249790-cb9c9513/train.jsonl

@@ -45,3 +45,4 @@
 {"step": 119500, "loss": 0.5577743649482727, "lr": 0.0003, "grad_norm": 0.13060694371417733, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17487473060, "tokens_per_second": 151600.22361494292, "data_wait_s": 1.4679000742034987e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 119600, "loss": 0.6463970169425011, "lr": 0.0003, "grad_norm": 0.15078117921742035, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17502101396, "tokens_per_second": 150354.02502220837, "data_wait_s": 9.141999726125505e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 119700, "loss": 0.6455818898975849, "lr": 0.0003, "grad_norm": 0.13960348146462503, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17516653535, "tokens_per_second": 149283.06221811823, "data_wait_s": 1.7284000023209956e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 119800, "loss": 0.5821646936237812, "lr": 0.0003, "grad_norm": 0.15459538561565184, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17531207737, "tokens_per_second": 148681.6515178564, "data_wait_s": 1.4895000276737846e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790519810.cad0b19a-59f2-43f0-8fb4-2a94097843e2-tjh7c.1336.0.1790519810984249790-cb9c9513

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b308ba420a3893e64c197e8b59ec4954507722541a811d593cd4089afdaf2f46
-size 99711
+oid sha256:f6620b62c3e609f5d54d0c17a6762d533e2bd0837890f81a06f750e8ba031cc9
+size 101829