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": 75090.32698031457,
-      "required_pages_per_second_upper_estimate": 0.006508639227959877
+      "required_token_positions_per_second": 75566.22687872914,
+      "required_pages_per_second_upper_estimate": 0.006549889024997186
     },
     {
       "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": 9386.29087253932,
-      "required_pages_per_second_upper_estimate": 0.0037443317666105477
+      "required_token_positions_per_second": 9445.778359841142,
+      "required_pages_per_second_upper_estimate": 0.003768062214712439
     },
     {
       "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": 18772.58174507864,
-      "required_pages_per_second_upper_estimate": 0.004342706400569319
+      "required_token_positions_per_second": 18891.556719682285,
+      "required_pages_per_second_upper_estimate": 0.004370229167055833
     },
     {
       "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": 28158.872617617963,
-      "required_pages_per_second_upper_estimate": 0.001714727604226118
+      "required_token_positions_per_second": 28337.33507952343,
+      "required_pages_per_second_upper_estimate": 0.0017255950318359858
     },
     {
       "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": 28158.872617617963,
-      "required_pages_per_second_upper_estimate": 0.0021677479628525783
+      "required_token_positions_per_second": 28337.33507952343,
+      "required_pages_per_second_upper_estimate": 0.002181486497185774
     },
     {
       "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": 28158.872617617963,
-      "required_pages_per_second_upper_estimate": 0.3496389562265538
+      "required_token_positions_per_second": 28337.33507952343,
+      "required_pages_per_second_upper_estimate": 0.3518548627279813
     }
   ],
   "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": 0.41852917088277053,
-    "producer_p95_seconds": 0.19930798500081437,
-    "gpu_update_seconds": 0.9818574903706078,
-    "producer_batches_per_second": 2.389319716689709,
-    "gpu_batches_per_second": 1.01847774224602,
-    "producer_headroom": 2.345971460721969,
+    "producer_batch_seconds": 0.07221680503906924,
+    "producer_p95_seconds": 0.19432173800032615,
+    "gpu_update_seconds": 0.9756739623684112,
+    "producer_batches_per_second": 13.847192484616297,
+    "gpu_batches_per_second": 1.0249325477258184,
+    "producer_headroom": 13.510345159143668,
     "sustainable": true,
-    "buffer_seconds": 60.87516440297768,
+    "buffer_seconds": 60.491785666841494,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 344843.3217181816,
+    "generation_bytes_per_second": 360232.771950771,
     "upload_bytes_per_second": 5874755.128837641,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 73.99534929904065,
+    "estimated_full_shards_per_hour": 77.29756707088802,
     "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

@@ -64,3 +64,4 @@
 {"step": 46400, "loss": 0.6492524016648531, "lr": 0.0003, "grad_norm": 0.14303506288905476, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6789668132, "tokens_per_second": 150701.92041407048, "data_wait_s": 1.6092999430838972e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 46500, "loss": 0.6424358170479536, "lr": 0.0003, "grad_norm": 0.13425616800438053, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6804374052, "tokens_per_second": 149083.97143347966, "data_wait_s": 1.0923999980150256e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 46600, "loss": 0.6715296972543001, "lr": 0.0003, "grad_norm": 0.13524944828831933, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6819165140, "tokens_per_second": 150721.4699803789, "data_wait_s": 1.1185999937879387e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 46700, "loss": 0.7140990644693375, "lr": 0.0003, "grad_norm": 0.14737092949938116, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6833976873, "tokens_per_second": 151543.23811024037, "data_wait_s": 9.436000254936516e-06, "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 sha256:3dd11ee2d3bb82614ede65d67466abacde3c76f98caaf418c94c90ce3ad6703b
-size 140736
+oid sha256:0a8a0e2f4d9af2a1a0d250526d40e99433adb21260e10af83af679e409706d1b
+size 142854