Update TensorBoard training metrics

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 598315.063954628,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 78582.4952529274,
-      "required_pages_per_second_upper_estimate": 0.005436519358107573
+      "required_token_positions_per_second": 78157.97306881017,
+      "required_pages_per_second_upper_estimate": 0.0054071499283831655
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 8533.957213384612,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9822.811906615925,
-      "required_pages_per_second_upper_estimate": 0.004112606221662091
+      "required_token_positions_per_second": 9769.746633601271,
+      "required_pages_per_second_upper_estimate": 0.0040903888999797655
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 380144.35761751566,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19645.62381323185,
-      "required_pages_per_second_upper_estimate": 0.0040916653503562
+      "required_token_positions_per_second": 19539.493267202542,
+      "required_pages_per_second_upper_estimate": 0.004069561156469009
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 43742.16556775435,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29468.435719847774,
-      "required_pages_per_second_upper_estimate": 0.0017665877257088652
+      "required_token_positions_per_second": 29309.239900803812,
+      "required_pages_per_second_upper_estimate": 0.0017570441794351205
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 22018.86276523721,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29468.435719847774,
-      "required_pages_per_second_upper_estimate": 0.00229104125926788
+      "required_token_positions_per_second": 29309.239900803812,
+      "required_pages_per_second_upper_estimate": 0.002278664484565618
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 910315.184299418,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 23,
-      "required_token_positions_per_second": 29468.435719847774,
-      "required_pages_per_second_upper_estimate": 0.37798945267310285
+      "required_token_positions_per_second": 29309.239900803812,
+      "required_pages_per_second_upper_estimate": 0.37594745963755993
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,18 +158,18 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 1.4968346825000083,
-    "producer_p95_seconds": 2.975958035000019,
-    "gpu_update_seconds": 0.9382242159999805,
-    "producer_batches_per_second": 0.6680764493843792,
-    "gpu_batches_per_second": 1.0658433058394017,
-    "producer_headroom": 0.6268055029517098,
-    "sustainable": false,
-    "buffer_seconds": 60.04634982399875,
-    "buffer_drain_seconds": 160.89827234571717,
+    "producer_batch_seconds": 0.043296239296877204,
+    "producer_p95_seconds": 0.10546573299995998,
+    "gpu_update_seconds": 0.9433202666999818,
+    "producer_batches_per_second": 23.096694221941956,
+    "gpu_batches_per_second": 1.0600853552084712,
+    "producer_headroom": 21.787579753330217,
+    "sustainable": true,
+    "buffer_seconds": 60.372497068798836,
+    "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 6,
-    "max_pending_pages": 1511,
+    "upload_pages_per_commit": 22,
+    "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,
     "upload_sustainable": null,

metrics/void-byte-v1/1791399463488568880-054ba646/train.jsonl

@@ -0,0 +1 @@
+{"step": 885100, "loss": 0.5562257170677185, "lr": 3e-05, "grad_norm": 0.32036708267067454, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 129553907620, "tokens_per_second": 153137.68317948302, "data_wait_s": 2.0897999888802588e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791399463.60f38c85-66c9-4fe6-8e58-7b3a56dc2cee-wtkw8.346.0.1791399463488568880-054ba646

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8ba665f58097c84a5f247fea7b0019afecdd7cc77805b445ec9c56e337bc5be8
-size 165
+oid sha256:4f103a53bacd5f4b1bfa2413e9229b41b55acbca07479e211d135f428cf9d373
+size 2283