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": 77457.8731288884,
-      "required_pages_per_second_upper_estimate": 0.005358715390083711
+      "required_token_positions_per_second": 77454.95328031323,
+      "required_pages_per_second_upper_estimate": 0.005358513388184299
     },
     {
       "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": 9682.23414111105,
-      "required_pages_per_second_upper_estimate": 0.004053749246842762
+      "required_token_positions_per_second": 9681.869160039154,
+      "required_pages_per_second_upper_estimate": 0.00405359643688963
     },
     {
       "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": 19364.4682822221,
-      "required_pages_per_second_upper_estimate": 0.00403310806781683
+      "required_token_positions_per_second": 19363.738320078308,
+      "required_pages_per_second_upper_estimate": 0.004032956035952675
     },
     {
       "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": 29046.70242333315,
-      "required_pages_per_second_upper_estimate": 0.0017413054585322715
+      "required_token_positions_per_second": 29045.607480117465,
+      "required_pages_per_second_upper_estimate": 0.0017412398183583725
     },
     {
       "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": 29046.70242333315,
-      "required_pages_per_second_upper_estimate": 0.0022582533504725963
+      "required_token_positions_per_second": 29045.607480117465,
+      "required_pages_per_second_upper_estimate": 0.0022581682234537226
     },
     {
       "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": 29046.70242333315,
-      "required_pages_per_second_upper_estimate": 0.37257991076734714
+      "required_token_positions_per_second": 29045.607480117465,
+      "required_pages_per_second_upper_estimate": 0.37256586601143477
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.051734507203111235,
-    "producer_p95_seconds": 0.14023769099958372,
-    "gpu_update_seconds": 0.9518464298305485,
-    "producer_batches_per_second": 19.32945830669595,
-    "gpu_batches_per_second": 1.05058964204764,
-    "producer_headroom": 18.39867587978698,
+    "producer_batch_seconds": 0.05044044935937464,
+    "producer_p95_seconds": 0.13483983799960697,
+    "gpu_update_seconds": 0.9518823119442702,
+    "producer_batches_per_second": 19.82535866949299,
+    "gpu_batches_per_second": 1.0505500390667486,
+    "producer_headroom": 18.871408245441366,
     "sustainable": true,
-    "buffer_seconds": 60.9181715091551,
+    "buffer_seconds": 60.92046796443329,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 80,
+    "upload_pages_per_commit": 87,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791254470898546162-7724d9f6/train.jsonl

@@ -29,3 +29,4 @@
 {"step": 762900, "loss": 0.5140604376792908, "lr": 3e-05, "grad_norm": 0.2409726487696894, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111666008715, "tokens_per_second": 151746.17577388752, "data_wait_s": 1.589900011822465e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 763000, "loss": 0.577847421169281, "lr": 3e-05, "grad_norm": 0.2346762175215377, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111680576336, "tokens_per_second": 152632.09971097927, "data_wait_s": 1.1066000297432765e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 763100, "loss": 0.5308964848518372, "lr": 3e-05, "grad_norm": 0.24813254160996756, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111695162754, "tokens_per_second": 152733.97305566468, "data_wait_s": 9.846000011748401e-06, "gpu_peak_gb": 89.37899398803711}
+{"step": 763200, "loss": 0.5169625878334045, "lr": 3e-05, "grad_norm": 0.22978181003732107, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111709830959, "tokens_per_second": 153816.96861438148, "data_wait_s": 1.4072999874770176e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791254470.1e0c9aa3-e42e-441e-9e17-2e25895f438d-9bhv5.366.0.1791254470898546162-7724d9f6

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8c949cce154b9aa69fc44bab8f053c9e96d5cd9b298a1787840b3b075a8dbff8
-size 65823
+oid sha256:ade381079d900be4ee297dd85736c3f3ab7c68472faf327ac0903f8a5e8e2558
+size 67941