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": 75708.50481288647,
-      "required_pages_per_second_upper_estimate": 0.0065622213157298164
+      "required_token_positions_per_second": 76003.97837554659,
+      "required_pages_per_second_upper_estimate": 0.006587832215270297
     },
     {
       "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": 9463.56310161081,
-      "required_pages_per_second_upper_estimate": 0.0037751568141099444
+      "required_token_positions_per_second": 9500.497296943324,
+      "required_pages_per_second_upper_estimate": 0.003789890416843515
     },
     {
       "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": 18927.12620322162,
-      "required_pages_per_second_upper_estimate": 0.0043784575410711285
+      "required_token_positions_per_second": 19000.994593886648,
+      "required_pages_per_second_upper_estimate": 0.00439554569321221
     },
     {
       "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": 28390.689304832424,
-      "required_pages_per_second_upper_estimate": 0.0017288440242293173
+      "required_token_positions_per_second": 28501.491890829973,
+      "required_pages_per_second_upper_estimate": 0.0017355913203803276
     },
     {
       "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": 28390.689304832424,
-      "required_pages_per_second_upper_estimate": 0.002185593853144009
+      "required_token_positions_per_second": 28501.491890829973,
+      "required_pages_per_second_upper_estimate": 0.0021941237429352893
     },
     {
       "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": 28390.689304832424,
-      "required_pages_per_second_upper_estimate": 0.3525173436412137
+      "required_token_positions_per_second": 28501.491890829973,
+      "required_pages_per_second_upper_estimate": 0.3538931409269028
     }
   ],
   "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": 0.1230204807266091,
-    "producer_p95_seconds": 0.3532341780010029,
-    "gpu_update_seconds": 0.9738403919377184,
-    "producer_batches_per_second": 8.12872778657336,
-    "gpu_batches_per_second": 1.0268623157129784,
-    "producer_headroom": 7.916083453631624,
+    "producer_batch_seconds": 0.2382218273752983,
+    "producer_p95_seconds": 0.3580973400021321,
+    "gpu_update_seconds": 0.9700544836705698,
+    "producer_batches_per_second": 4.197768151717621,
+    "gpu_batches_per_second": 1.0308699323940238,
+    "producer_headroom": 4.072063816983198,
     "sustainable": true,
-    "buffer_seconds": 60.378104300138546,
+    "buffer_seconds": 60.14337798757533,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 86,
+    "upload_pages_per_commit": 81,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 518657.1922661141,
+    "generation_bytes_per_second": 507108.8732165754,
     "upload_bytes_per_second": 3935103.367630573,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 111.29175973880355,
+    "estimated_full_shards_per_hour": 108.81375930188129,
     "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

@@ -326,3 +326,4 @@
 {"step": 72600, "loss": 0.6149622984230518, "lr": 0.0003, "grad_norm": 0.13127798112473363, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 10625047722, "tokens_per_second": 149422.48085558575, "data_wait_s": 1.017199974739924e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 72700, "loss": 0.6209270842373371, "lr": 0.0003, "grad_norm": 0.13544607122665395, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 10639721542, "tokens_per_second": 150185.50780856414, "data_wait_s": 1.168000017059967e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 72800, "loss": 0.6211061533540487, "lr": 0.0003, "grad_norm": 0.15273745097921554, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 10654408420, "tokens_per_second": 150021.76071150662, "data_wait_s": 9.208997653331608e-06, "gpu_peak_gb": 85.46427917480469}
+{"step": 72900, "loss": 0.6699622832238674, "lr": 0.0003, "grad_norm": 0.1450948861588816, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 10668920705, "tokens_per_second": 148963.6484958983, "data_wait_s": 1.0960997315123677e-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 sha256:4927d0baa0821b31a860c04f7d304b3af17b24a3b32235da7ef3091f1ae9996c
-size 699567
+oid sha256:915f71dd0e2658de926e9f494460ac386897949a20c96fb96f3a4fde51ec9a50
+size 701685