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": 76447.35125617971,
-      "required_pages_per_second_upper_estimate": 0.006626262652845284
+      "required_token_positions_per_second": 76452.41900980516,
+      "required_pages_per_second_upper_estimate": 0.006626701912885434
     },
     {
       "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": 9555.918907022464,
-      "required_pages_per_second_upper_estimate": 0.003811998925730997
+      "required_token_positions_per_second": 9556.552376225645,
+      "required_pages_per_second_upper_estimate": 0.003812251626067355
     },
     {
       "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": 19111.837814044928,
-      "required_pages_per_second_upper_estimate": 0.00442118732142178
+      "required_token_positions_per_second": 19113.10475245129,
+      "required_pages_per_second_upper_estimate": 0.0044214804053247375
     },
     {
       "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": 28667.75672106739,
-      "required_pages_per_second_upper_estimate": 0.0017457159762176387
+      "required_token_positions_per_second": 28669.657128676932,
+      "required_pages_per_second_upper_estimate": 0.0017458317010703916
     },
     {
       "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": 28667.75672106739,
-      "required_pages_per_second_upper_estimate": 0.002206923269817473
+      "required_token_positions_per_second": 28669.657128676932,
+      "required_pages_per_second_upper_estimate": 0.0022070695684559207
     },
     {
       "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": 28667.75672106739,
-      "required_pages_per_second_upper_estimate": 0.35595759366586033
+      "required_token_positions_per_second": 28669.657128676932,
+      "required_pages_per_second_upper_estimate": 0.35598119036811565
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 690670,
-    "producer_batch_seconds": 0.05250168803906874,
-    "producer_p95_seconds": 0.1334488949999013,
-    "gpu_update_seconds": 0.9644284437394436,
-    "producer_batches_per_second": 19.047006626831834,
-    "gpu_batches_per_second": 1.0368835619599028,
-    "producer_headroom": 18.369474959010297,
+    "producer_batch_seconds": 0.051747158281244765,
+    "producer_p95_seconds": 0.13375798200013378,
+    "gpu_update_seconds": 0.9643645152751054,
+    "producer_batches_per_second": 19.324732665802053,
+    "gpu_batches_per_second": 1.0369522977675396,
+    "producer_headroom": 18.636086450077194,
     "sustainable": true,
-    "buffer_seconds": 60.758991955584946,
+    "buffer_seconds": 60.754964462331635,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 57,
+    "upload_pages_per_commit": 62,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790696019647648547-3fdea0a8/train.jsonl

@@ -7,3 +7,4 @@
 {"step": 275700, "loss": 0.5936791710555553, "lr": 0.0003, "grad_norm": 0.15593649290911388, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40354109564, "tokens_per_second": 149561.37570848665, "data_wait_s": 1.07790001493413e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 275800, "loss": 0.5612323526293039, "lr": 0.0003, "grad_norm": 0.14636701453080406, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40368593068, "tokens_per_second": 149577.26139316798, "data_wait_s": 9.232999900632421e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 275900, "loss": 0.5655969046056271, "lr": 0.0003, "grad_norm": 0.14367211251604423, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40383055030, "tokens_per_second": 149426.43140356545, "data_wait_s": 1.1692999805745785e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 276000, "loss": 0.6486665718257427, "lr": 0.0003, "grad_norm": 0.146840384138437, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40397759181, "tokens_per_second": 152216.3817171238, "data_wait_s": 1.513699999122764e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790696019.3e6d470a-8cc7-4d7e-a224-51a0f9a2ce55-85btc.22407.0.1790696019647648547-3fdea0a8

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9af842fe12d071899464adcf18a26d5be8d866f79ec85cc1c41ea07a3a26c515
-size 19227
+oid sha256:86542276e4880219a1abfe6a8a6f15bcc982bad1925ed2b4d67920c5243e49a9
+size 21345