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": 76949.51291318302,
-      "required_pages_per_second_upper_estimate": 0.005323545850803001
+      "required_token_positions_per_second": 76888.59938250446,
+      "required_pages_per_second_upper_estimate": 0.005319331711411794
     },
     {
       "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": 9618.689114147877,
-      "required_pages_per_second_upper_estimate": 0.004027144271024339
+      "required_token_positions_per_second": 9611.074922813057,
+      "required_pages_per_second_upper_estimate": 0.0040239563681148454
     },
     {
       "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": 19237.378228295755,
-      "required_pages_per_second_upper_estimate": 0.004006638561174079
+      "required_token_positions_per_second": 19222.149845626114,
+      "required_pages_per_second_upper_estimate": 0.00400346689066345
     },
     {
       "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": 28856.06734244363,
-      "required_pages_per_second_upper_estimate": 0.001729877176515859
+      "required_token_positions_per_second": 28833.22476843917,
+      "required_pages_per_second_upper_estimate": 0.0017285077990828825
     },
     {
       "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": 28856.06734244363,
-      "required_pages_per_second_upper_estimate": 0.0022434323114485398
+      "required_token_positions_per_second": 28833.22476843917,
+      "required_pages_per_second_upper_estimate": 0.0022416564017935575
     },
     {
       "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": 28856.06734244363,
-      "required_pages_per_second_upper_estimate": 0.37013464863769874
+      "required_token_positions_per_second": 28833.22476843917,
+      "required_pages_per_second_upper_estimate": 0.36984164862481456
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.04942711695312241,
-    "producer_p95_seconds": 0.14132986799995706,
-    "gpu_update_seconds": 0.9581347198803242,
-    "producer_batches_per_second": 20.231809210082364,
-    "gpu_batches_per_second": 1.0436945653372263,
-    "producer_headroom": 19.384798850174427,
+    "producer_batch_seconds": 0.05199078326562656,
+    "producer_p95_seconds": 0.14265257100032613,
+    "gpu_update_seconds": 0.9588937838914044,
+    "producer_batches_per_second": 19.23417839063688,
+    "gpu_batches_per_second": 1.0428683727010695,
+    "producer_headroom": 18.443534097040082,
     "sustainable": true,
-    "buffer_seconds": 60.36248735246042,
+    "buffer_seconds": 60.41030838515848,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 75,
+    "upload_pages_per_commit": 81,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791209746770022608-f3fe5fb7/train.jsonl

@@ -21,3 +21,4 @@
 {"step": 722100, "loss": 0.5536484122276306, "lr": 5e-05, "grad_norm": 0.21087483701386997, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105695631662, "tokens_per_second": 151489.05163858933, "data_wait_s": 8.944000001065433e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 722200, "loss": 0.5455427169799805, "lr": 5e-05, "grad_norm": 0.21476824125879201, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105710086105, "tokens_per_second": 150554.98317487695, "data_wait_s": 1.0819000181072624e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 722300, "loss": 0.502178966999054, "lr": 5e-05, "grad_norm": 0.21813273108140901, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105724561000, "tokens_per_second": 150706.10889914926, "data_wait_s": 1.2273000265849987e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 722400, "loss": 0.5104002952575684, "lr": 5e-05, "grad_norm": 0.2041116847258441, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105739363686, "tokens_per_second": 153919.0252666146, "data_wait_s": 1.1156999789818656e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791209746.ebd20fd8-0561-42d8-bf2b-56d08e7106ef-6qqk2.329.0.1791209746770022608-f3fe5fb7

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d3f69d551251ee9008925bcea1a7272d192b92387021062e3b7fd68304aa430e
-size 48879
+oid sha256:e8cc0ec46097c8f08d42f14a5ee84ed776be8d6cc031b71266e4adf35aa3dc77
+size 50997