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": 76893.42526090816,
-      "required_pages_per_second_upper_estimate": 0.005319665576877342
+      "required_token_positions_per_second": 76732.20180767066,
+      "required_pages_per_second_upper_estimate": 0.005308511764292421
     },
     {
       "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": 9611.67815761352,
-      "required_pages_per_second_upper_estimate": 0.0040242089299288245
+      "required_token_positions_per_second": 9591.525225958832,
+      "required_pages_per_second_upper_estimate": 0.004015771318286075
     },
     {
       "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": 19223.35631522704,
-      "required_pages_per_second_upper_estimate": 0.00400371816646458
+      "required_token_positions_per_second": 19183.050451917665,
+      "required_pages_per_second_upper_estimate": 0.003995323518074332
     },
     {
       "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": 28835.03447284056,
-      "required_pages_per_second_upper_estimate": 0.0017286162880984872
+      "required_token_positions_per_second": 28774.5756778765,
+      "required_pages_per_second_upper_estimate": 0.0017249918756556265
     },
     {
       "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": 28835.03447284056,
-      "required_pages_per_second_upper_estimate": 0.0022417970983507176
+      "required_token_positions_per_second": 28774.5756778765,
+      "required_pages_per_second_upper_estimate": 0.002237096692972396
     },
     {
       "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": 28835.03447284056,
-      "required_pages_per_second_upper_estimate": 0.36986486156976645
+      "required_token_positions_per_second": 28774.5756778765,
+      "required_pages_per_second_upper_estimate": 0.3690893610635638
     }
   ],
   "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": 0.05245892740626701,
-    "producer_p95_seconds": 0.12827070399998775,
-    "gpu_update_seconds": 0.9588336031309892,
-    "producer_batches_per_second": 19.062532336117396,
-    "gpu_batches_per_second": 1.0429338278660505,
-    "producer_headroom": 18.277796564640436,
+    "producer_batch_seconds": 0.04453031145314412,
+    "producer_p95_seconds": 0.12385747000007541,
+    "gpu_update_seconds": 0.9608482262088517,
+    "producer_batches_per_second": 22.456613649608634,
+    "gpu_batches_per_second": 1.0407470948306026,
+    "producer_headroom": 21.577397391883945,
     "sustainable": true,
-    "buffer_seconds": 60.40651699725232,
+    "buffer_seconds": 60.533438251157655,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 66,
+    "upload_pages_per_commit": 69,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791300814123050567-72be911d/train.jsonl

@@ -12,3 +12,4 @@
 {"step": 801200, "loss": 0.5780178308486938, "lr": 3e-05, "grad_norm": 0.2919761416088437, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117272170109, "tokens_per_second": 152201.3480068655, "data_wait_s": 2.256599964312045e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 801300, "loss": 0.5488638877868652, "lr": 3e-05, "grad_norm": 0.27684681061972655, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117286883998, "tokens_per_second": 153382.37674826293, "data_wait_s": 9.282000064558815e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 801400, "loss": 0.5533844232559204, "lr": 3e-05, "grad_norm": 0.2832121404691586, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117301622732, "tokens_per_second": 153554.14737391332, "data_wait_s": 1.5303000054700533e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 801500, "loss": 0.5405973792076111, "lr": 3e-05, "grad_norm": 0.2594121739967866, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117316172213, "tokens_per_second": 151355.2211082398, "data_wait_s": 1.9609999981184956e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791300814.0c9d201b-2dbc-40fb-b739-4ee2350a0f54-7b2tx.840.0.1791300814123050567-72be911d

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6465be9c6804ed93b2d8bd982902c87f727c9bebbcba84ada30c873e4e344d3e
-size 29817
+oid sha256:9c7ba86211be09889d54e6928469d3fb4738cb49e15386a9b05ee9ca0f470a2d
+size 31935