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": 75216.9209802901,
-      "required_pages_per_second_upper_estimate": 0.0065196120750281965
+      "required_token_positions_per_second": 75157.22568400606,
+      "required_pages_per_second_upper_estimate": 0.006514437838042641
     },
     {
       "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": 9402.115122536263,
-      "required_pages_per_second_upper_estimate": 0.003750644296527949
+      "required_token_positions_per_second": 9394.653210500757,
+      "required_pages_per_second_upper_estimate": 0.0037476676282514588
     },
     {
       "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": 18804.230245072526,
-      "required_pages_per_second_upper_estimate": 0.004350027724048328
+      "required_token_positions_per_second": 18789.306421001515,
+      "required_pages_per_second_upper_estimate": 0.004346575359999832
     },
     {
       "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": 28206.34536760879,
-      "required_pages_per_second_upper_estimate": 0.0017176184456302891
+      "required_token_positions_per_second": 28183.95963150227,
+      "required_pages_per_second_upper_estimate": 0.0017162552717502827
     },
     {
       "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": 28206.34536760879,
-      "required_pages_per_second_upper_estimate": 0.002171402546559855
+      "required_token_positions_per_second": 28183.95963150227,
+      "required_pages_per_second_upper_estimate": 0.0021696792306266912
     },
     {
       "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": 28206.34536760879,
-      "required_pages_per_second_upper_estimate": 0.35022840890036616
+      "required_token_positions_per_second": 28183.95963150227,
+      "required_pages_per_second_upper_estimate": 0.34995045297816246
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,24 +158,24 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 690670,
-    "producer_batch_seconds": 0.11901390506253051,
-    "producer_p95_seconds": 0.3537828329972399,
-    "gpu_update_seconds": 0.9802049730182354,
-    "producer_batches_per_second": 8.402379532665488,
-    "gpu_batches_per_second": 1.0201947832613132,
-    "producer_headroom": 8.236054203105349,
+    "producer_batch_seconds": 0.11783737476574174,
+    "producer_p95_seconds": 0.3697225180003443,
+    "gpu_update_seconds": 0.9809835225954834,
+    "producer_batches_per_second": 8.486271880954403,
+    "gpu_batches_per_second": 1.0193851139866272,
+    "producer_headroom": 8.324892883481649,
     "sustainable": true,
-    "buffer_seconds": 60.772708327130594,
+    "buffer_seconds": 60.82097840091997,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 83,
+    "upload_pages_per_commit": 82,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 321308.71778962703,
+    "generation_bytes_per_second": 321072.0685865321,
     "upload_bytes_per_second": 4134798.041822171,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 68.94537115351304,
+    "estimated_full_shards_per_hour": 68.89459174344036,
     "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
   },
   "upload_measurements": [

metrics/void-byte-v1/1790738827394728502-c744ae20/train.jsonl

@@ -292,3 +292,4 @@
 {"step": 344200, "loss": 0.5796448960900307, "lr": 0.0003, "grad_norm": 0.15931724983149692, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 50381281804, "tokens_per_second": 146848.80195492788, "data_wait_s": 1.6038997273426503e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 344300, "loss": 0.6537228301167488, "lr": 0.0003, "grad_norm": 0.18833676195820212, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 50395982680, "tokens_per_second": 149616.36619675183, "data_wait_s": 1.4023000403540209e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 344400, "loss": 0.5523795243352652, "lr": 0.0003, "grad_norm": 0.15344267455651656, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 50410762680, "tokens_per_second": 149709.6909339595, "data_wait_s": 1.4655997802037746e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 344500, "loss": 0.5823973026126623, "lr": 0.0003, "grad_norm": 0.16255057540814566, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 50425496297, "tokens_per_second": 150492.55814300498, "data_wait_s": 1.4551998901879415e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790738827.1a46e860-412e-40b5-8c01-42d24543f8c9-dfnph.319.0.1790738827394728502-c744ae20

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7e6a695175119099e19c032e149a50fefc940eefd3ec661ae053e47f09e8d976
-size 626772
+oid sha256:6ce0fd817eb5632158e4d4ded7a2fc4b8a8c720423c21f0775db7db7aaf85a01
+size 628890