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": 75585.30080337779,
-      "required_pages_per_second_upper_estimate": 0.006551542304443311
+      "required_token_positions_per_second": 75768.92635854639,
+      "required_pages_per_second_upper_estimate": 0.0065674585021706115
     },
     {
       "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": 9448.162600422223,
-      "required_pages_per_second_upper_estimate": 0.003769013323927806
+      "required_token_positions_per_second": 9471.115794818299,
+      "required_pages_per_second_upper_estimate": 0.003778169696353239
     },
     {
       "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": 18896.325200844447,
-      "required_pages_per_second_upper_estimate": 0.004371332271250298
+      "required_token_positions_per_second": 18942.231589636598,
+      "required_pages_per_second_upper_estimate": 0.004381951906372513
     },
     {
       "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": 28344.48780126667,
-      "required_pages_per_second_upper_estimate": 0.0017260305950627186
+      "required_token_positions_per_second": 28413.347384454897,
+      "required_pages_per_second_upper_estimate": 0.0017302237824005696
     },
     {
       "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": 28344.48780126667,
-      "required_pages_per_second_upper_estimate": 0.00218203713350557
+      "required_token_positions_per_second": 28413.347384454897,
+      "required_pages_per_second_upper_estimate": 0.002187338134834926
     },
     {
       "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": 28344.48780126667,
-      "required_pages_per_second_upper_estimate": 0.3519436755934126
+      "required_token_positions_per_second": 28413.347384454897,
+      "required_pages_per_second_upper_estimate": 0.3527986811584104
     }
   ],
   "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": 0.050977172562514994,
-    "producer_p95_seconds": 0.14861546000020098,
-    "gpu_update_seconds": 0.9754277513797394,
-    "producer_batches_per_second": 19.616623475412784,
-    "gpu_batches_per_second": 1.025191254386092,
-    "producer_headroom": 19.134598926284898,
+    "producer_batch_seconds": 0.048770584289059116,
+    "producer_p95_seconds": 0.14079160400024193,
+    "gpu_update_seconds": 0.9730638078611736,
+    "producer_batches_per_second": 20.504162797662723,
+    "gpu_batches_per_second": 1.0276818353752495,
+    "producer_headroom": 19.951858728899104,
     "sustainable": true,
-    "buffer_seconds": 60.47652058554384,
+    "buffer_seconds": 60.329956087392766,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 72,
+    "upload_pages_per_commit": 86,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -16,3 +16,4 @@
 {"step": 116600, "loss": 0.6223240718245506, "lr": 0.0003, "grad_norm": 0.14137559062344082, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17062709580, "tokens_per_second": 151335.11674431755, "data_wait_s": 1.1185000403202139e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 116700, "loss": 0.6237425673753023, "lr": 0.0003, "grad_norm": 0.13083552124961959, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17077478076, "tokens_per_second": 151831.34323416388, "data_wait_s": 1.7583000044396613e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 116800, "loss": 0.6324888169765472, "lr": 0.0003, "grad_norm": 0.13479138816512448, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17092217704, "tokens_per_second": 151539.56402974177, "data_wait_s": 1.4374999864230631e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 116900, "loss": 0.6173514053225517, "lr": 0.0003, "grad_norm": 0.13649673275466037, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17106677430, "tokens_per_second": 148989.21334799164, "data_wait_s": 1.3545999991038116e-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 sha256:7cf10f096b230700c5cf5f05eb277c3c2ea7356c5347da73f1f404beb58fe031
-size 38289
+oid sha256:e3b53ddf9af85322d6885b53c5ed810cb7875e7e66c063918aea15e968d3c2d9
+size 40407