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": 75139.44171383239,
-      "required_pages_per_second_upper_estimate": 0.006512896368581047
+      "required_token_positions_per_second": 75325.43923111251,
+      "required_pages_per_second_upper_estimate": 0.00652901816197249
     },
     {
       "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": 9392.430214229049,
-      "required_pages_per_second_upper_estimate": 0.0037467808418019184
+      "required_token_positions_per_second": 9415.679903889064,
+      "required_pages_per_second_upper_estimate": 0.003756055490621136
     },
     {
       "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": 18784.860428458098,
-      "required_pages_per_second_upper_estimate": 0.004345546857871709
+      "required_token_positions_per_second": 18831.359807778128,
+      "required_pages_per_second_upper_estimate": 0.004356303670916273
     },
     {
       "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": 28177.290642687145,
-      "required_pages_per_second_upper_estimate": 0.0017158491653209216
+      "required_token_positions_per_second": 28247.039711667192,
+      "required_pages_per_second_upper_estimate": 0.0017200965176767235
     },
     {
       "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": 28177.290642687145,
-      "required_pages_per_second_upper_estimate": 0.0021691658334067605
+      "required_token_positions_per_second": 28247.039711667192,
+      "required_pages_per_second_upper_estimate": 0.002174535309814625
     },
     {
       "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": 28177.290642687145,
-      "required_pages_per_second_upper_estimate": 0.3498676464567484
+      "required_token_positions_per_second": 28247.039711667192,
+      "required_pages_per_second_upper_estimate": 0.35073369645836316
     }
   ],
   "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": 0.03887681627354311,
-    "producer_p95_seconds": 0.13568076900082815,
-    "gpu_update_seconds": 0.9812157013462005,
-    "producer_batches_per_second": 25.7222709021194,
-    "gpu_batches_per_second": 1.019143903453673,
-    "producer_headroom": 25.239096083440053,
+    "producer_batch_seconds": 0.038862487718802186,
+    "producer_p95_seconds": 0.12697996599854378,
+    "gpu_update_seconds": 0.9787928321770382,
+    "producer_batches_per_second": 25.73175467395997,
+    "gpu_batches_per_second": 1.0216666562379628,
+    "producer_headroom": 25.186057034210016,
     "sustainable": true,
-    "buffer_seconds": 60.83537348346443,
+    "buffer_seconds": 60.685155594976365,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 43,
+    "upload_pages_per_commit": 53,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790485060735632541-7579de36/train.jsonl

@@ -1,3 +1,4 @@
 {"step": 85100, "loss": 0.6527320481836796, "lr": 0.0003, "grad_norm": 0.13022073401510567, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12454100277, "tokens_per_second": 99305.6642030579, "data_wait_s": 1.0027999451267533e-05, "gpu_peak_gb": 85.46991729736328}
 {"step": 85200, "loss": 0.6027117408812046, "lr": 0.0003, "grad_norm": 0.1388974004150665, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12468770344, "tokens_per_second": 149461.1777327199, "data_wait_s": 1.0472000212757848e-05, "gpu_peak_gb": 85.46991729736328}
 {"step": 85300, "loss": 0.6186191104352474, "lr": 0.0003, "grad_norm": 0.13287940844382282, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12483365755, "tokens_per_second": 149447.52687704866, "data_wait_s": 1.2592001439770684e-05, "gpu_peak_gb": 85.46991729736328}
+{"step": 85400, "loss": 0.6122558545321226, "lr": 0.0003, "grad_norm": 0.1413275398458881, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12497887970, "tokens_per_second": 148746.17292177535, "data_wait_s": 6.886999472044408e-06, "gpu_peak_gb": 85.46991729736328}

runs/void-byte-v1/events.out.tfevents.1790485060.0b09bbb2-f535-42d2-bd00-cbae7fb94f51-r2qrq.28895.0.1790485060735632541-7579de36

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:57a0fe5e01d3964324ed43b29e2857f2ca4559ead6718b7d9197df8ed8510b7a
-size 6519
+oid sha256:e80b47d3283931ce3662e08737ec2699ab6d56cfca24a2615b2a1ae81dcc98f4
+size 8637