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": 75548.37368533963,
-      "required_pages_per_second_upper_estimate": 0.006548341555442688
+      "required_token_positions_per_second": 75133.13345379168,
+      "required_pages_per_second_upper_estimate": 0.006512349584588858
     },
     {
       "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": 9443.546710667453,
-      "required_pages_per_second_upper_estimate": 0.003767171976490926
+      "required_token_positions_per_second": 9391.64168172396,
+      "required_pages_per_second_upper_estimate": 0.003746466284396027
     },
     {
       "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": 18887.093421334906,
-      "required_pages_per_second_upper_estimate": 0.004369196661534536
+      "required_token_positions_per_second": 18783.28336344792,
+      "required_pages_per_second_upper_estimate": 0.004345182031637001
     },
     {
       "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": 28330.640132002358,
-      "required_pages_per_second_upper_estimate": 0.0017251873446576277
+      "required_token_positions_per_second": 28174.925045171884,
+      "required_pages_per_second_upper_estimate": 0.0017157051128435733
     },
     {
       "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": 28330.640132002358,
-      "required_pages_per_second_upper_estimate": 0.002180971101592801
+      "required_token_positions_per_second": 28174.925045171884,
+      "required_pages_per_second_upper_estimate": 0.0021689837231616427
     },
     {
       "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": 28330.640132002358,
-      "required_pages_per_second_upper_estimate": 0.35177173388631755
+      "required_token_positions_per_second": 28174.925045171884,
+      "required_pages_per_second_upper_estimate": 0.34983827365275444
     }
   ],
   "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": 688946,
-    "producer_batch_seconds": 0.14124316669534664,
-    "producer_p95_seconds": 0.4402986479981337,
-    "gpu_update_seconds": 0.97590452849559,
-    "producer_batches_per_second": 7.079988528980961,
-    "gpu_batches_per_second": 1.0246903982929094,
-    "producer_headroom": 6.909392867129351,
+    "producer_batch_seconds": 0.12603624823447035,
+    "producer_p95_seconds": 0.43727277699508704,
+    "gpu_update_seconds": 0.9812980852894166,
+    "producer_batches_per_second": 7.934225383634551,
+    "gpu_batches_per_second": 1.0190583422009507,
+    "producer_headroom": 7.785840177215272,
     "sustainable": true,
-    "buffer_seconds": 60.506080766726576,
+    "buffer_seconds": 60.84048128794383,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 270401.37842881365,
+    "generation_bytes_per_second": 287303.50039527926,
     "upload_bytes_per_second": 6345416.31829655,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 58.02184118889148,
+    "estimated_full_shards_per_hour": 61.64864310163291,
     "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-v1/1790434090210070261-a05fc3fb/train.jsonl

@@ -397,3 +397,4 @@
 {"step": 79700, "loss": 0.6050420720130205, "lr": 0.0003, "grad_norm": 0.1384838142148072, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 11663822473, "tokens_per_second": 150609.7271025089, "data_wait_s": 1.0958996426779777e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 79800, "loss": 0.6834732554852962, "lr": 0.0003, "grad_norm": 0.19023999182704254, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 11678369579, "tokens_per_second": 139857.67292984476, "data_wait_s": 1.2934004189446568e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 79900, "loss": 0.6089620068669319, "lr": 0.0003, "grad_norm": 0.14462929289110324, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 11692886215, "tokens_per_second": 146590.67159233542, "data_wait_s": 1.4140998246148229e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 80000, "loss": 0.5926032345741987, "lr": 0.0003, "grad_norm": 0.15032831750193268, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 11707359392, "tokens_per_second": 146935.6635780565, "data_wait_s": 1.720400177873671e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790434090.372bd512-57cc-4928-b44a-3b375a0fa071-5spg5.451.0.1790434090210070261-a05fc3fb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3f05328beccdcb6f9352c2dc008ae875b25b84e8b678279330d363ee93c295d1
-size 850728
+oid sha256:ac6b50c0efff366ef3c7a1395dea53ec2d2f04dcf1aa316bb57ec22616aecca8
+size 852846