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": 77289.54473033601,
-      "required_pages_per_second_upper_estimate": 0.005347070040896161
+      "required_token_positions_per_second": 77207.51516124002,
+      "required_pages_per_second_upper_estimate": 0.005341395044971277
     },
     {
       "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": 9661.193091292002,
-      "required_pages_per_second_upper_estimate": 0.0040449397986706105
+      "required_token_positions_per_second": 9650.939395155003,
+      "required_pages_per_second_upper_estimate": 0.004040646790219573
     },
     {
       "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": 19322.386182584003,
-      "required_pages_per_second_upper_estimate": 0.004024343476241811
+      "required_token_positions_per_second": 19301.878790310006,
+      "required_pages_per_second_upper_estimate": 0.0040200723272474405
     },
     {
       "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": 28983.579273876003,
-      "required_pages_per_second_upper_estimate": 0.0017375213221057838
+      "required_token_positions_per_second": 28952.81818546501,
+      "required_pages_per_second_upper_estimate": 0.0017356772418250072
     },
     {
       "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": 28983.579273876003,
-      "required_pages_per_second_upper_estimate": 0.0022533458032516942
+      "required_token_positions_per_second": 28952.81818546501,
+      "required_pages_per_second_upper_estimate": 0.002250954263931468
     },
     {
       "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": 28983.579273876003,
-      "required_pages_per_second_upper_estimate": 0.3717702347824682
+      "required_token_positions_per_second": 28952.81818546501,
+      "required_pages_per_second_upper_estimate": 0.3713756645690154
     }
   ],
   "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": 0.05531671475781286,
-    "producer_p95_seconds": 0.11484844599999633,
-    "gpu_update_seconds": 0.9539194500011318,
-    "producer_batches_per_second": 18.077718540918255,
-    "gpu_batches_per_second": 1.0483065420238717,
-    "producer_headroom": 17.244687327828004,
+    "producer_batch_seconds": 0.04637489773437142,
+    "producer_p95_seconds": 0.11100461700004871,
+    "gpu_update_seconds": 0.9549329472140968,
+    "producer_batches_per_second": 21.56338986940419,
+    "gpu_batches_per_second": 1.0471939447867842,
+    "producer_headroom": 20.59159143991674,
     "sustainable": true,
-    "buffer_seconds": 60.0969253500713,
+    "buffer_seconds": 60.1607756744881,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 31,
+    "upload_pages_per_commit": 43,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791463982175382982-d4a3b372/train.jsonl

@@ -1,2 +1,3 @@
 {"step": 925100, "loss": 0.5094915628433228, "lr": 3e-05, "grad_norm": 0.3323485047186109, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135408062366, "tokens_per_second": 152551.99412574965, "data_wait_s": 7.469000024684647e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 925200, "loss": 0.5638423562049866, "lr": 3e-05, "grad_norm": 0.3383034160205606, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135422770685, "tokens_per_second": 154311.20983017658, "data_wait_s": 1.4731999954165076e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 925300, "loss": 0.5204002857208252, "lr": 3e-05, "grad_norm": 0.3165123073884374, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135437541266, "tokens_per_second": 154976.86957509432, "data_wait_s": 1.3204000083533174e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791463982.2e076f0c-3fe2-4316-8f17-4e30b2e706e2-b5972.360.0.1791463982175382982-d4a3b372

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9acac6323db02a22a3da9eb21b6c2a949b48a31b3b10b5fab45eb59173172238
-size 4401
+oid sha256:b9da436e455da9154c65489f8e3cf528ac967e95961b51c593012f8325191a18
+size 6519