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": 75377.32570949705,
-      "required_pages_per_second_upper_estimate": 0.006533515550413785
+      "required_token_positions_per_second": 75707.42483765811,
+      "required_pages_per_second_upper_estimate": 0.006562127706214194
     },
     {
       "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": 9422.16571368713,
-      "required_pages_per_second_upper_estimate": 0.003758642777121083
+      "required_token_positions_per_second": 9463.428104707264,
+      "required_pages_per_second_upper_estimate": 0.0037751029618267365
     },
     {
       "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": 18844.33142737426,
-      "required_pages_per_second_upper_estimate": 0.00435930442681713
+      "required_token_positions_per_second": 18926.856209414527,
+      "required_pages_per_second_upper_estimate": 0.004378395082755587
     },
     {
       "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": 28266.497141061394,
-      "required_pages_per_second_upper_estimate": 0.001721281373041588
+      "required_token_positions_per_second": 28390.28431412179,
+      "required_pages_per_second_upper_estimate": 0.0017288193624198618
     },
     {
       "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": 28266.497141061394,
-      "required_pages_per_second_upper_estimate": 0.0021760331965910026
+      "required_token_positions_per_second": 28390.28431412179,
+      "required_pages_per_second_upper_estimate": 0.002185562675838018
     },
     {
       "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": 28266.497141061394,
-      "required_pages_per_second_upper_estimate": 0.350975292611612
+      "required_token_positions_per_second": 28390.28431412179,
+      "required_pages_per_second_upper_estimate": 0.3525123150120043
     }
   ],
   "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": 0.05134972500781387,
-    "producer_p95_seconds": 0.1276642670000001,
-    "gpu_update_seconds": 0.9781190736873112,
-    "producer_batches_per_second": 19.474300979174284,
-    "gpu_batches_per_second": 1.0223704116413987,
-    "producer_headroom": 19.04818523445785,
+    "producer_batch_seconds": 0.05179153343748588,
+    "producer_p95_seconds": 0.12582717800000864,
+    "gpu_update_seconds": 0.9738542838842735,
+    "producer_batches_per_second": 19.30817517127647,
+    "gpu_batches_per_second": 1.0268476676114653,
+    "producer_headroom": 18.803349104535556,
     "sustainable": true,
-    "buffer_seconds": 60.6433825686133,
+    "buffer_seconds": 60.37896560082496,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 59,
+    "upload_pages_per_commit": 60,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790606918076636751-3952f1c5/train.jsonl

@@ -8,3 +8,4 @@
 {"step": 195800, "loss": 0.6248968876898289, "lr": 0.0003, "grad_norm": 0.15336378557936461, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28657488126, "tokens_per_second": 151695.90750651533, "data_wait_s": 1.467199990656809e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 195900, "loss": 0.5964385215193033, "lr": 0.0003, "grad_norm": 0.1329017071350882, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28672190723, "tokens_per_second": 150429.23591048538, "data_wait_s": 9.922000117512653e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 196000, "loss": 0.5849446766078472, "lr": 0.0003, "grad_norm": 0.17091316453886277, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28686768471, "tokens_per_second": 148838.12571125917, "data_wait_s": 1.9231000123909325e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 196100, "loss": 0.5884930640459061, "lr": 0.0003, "grad_norm": 0.13476272340182135, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28701200508, "tokens_per_second": 147894.88826712858, "data_wait_s": 1.4421000059883227e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790606918.0a7f8bdc-23d3-4ecb-bd0c-4a4bf460026e-rl2cx.1192.0.1790606918076636751-3952f1c5

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b40b0f8e8aaa24f7479e0c3f1ad44290f745d4a1dfbf5227bd3338376f573c34
-size 21345
+oid sha256:1e6e901f21a3428abb0054b511ddd9043e9b59afbe8a48ec89bb6cb4cfe10572
+size 23463