Update TensorBoard training metrics

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 1062967.6168812835,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 75696.51773107261,
-      "required_pages_per_second_upper_estimate": 0.006561182305859173
+      "required_token_positions_per_second": 75739.94449385529,
+      "required_pages_per_second_upper_estimate": 0.006564946427593051
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 1206672.263377472,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9462.064716384077,
-      "required_pages_per_second_upper_estimate": 0.0037745590858401456
+      "required_token_positions_per_second": 9467.493061731911,
+      "required_pages_per_second_upper_estimate": 0.0037767245339603924
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 410042.7581134458,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 18924.129432768153,
-      "required_pages_per_second_upper_estimate": 0.004377764290968078
+      "required_token_positions_per_second": 18934.986123463823,
+      "required_pages_per_second_upper_estimate": 0.004380275795289291
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 2163486.5624068812,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28386.19414915223,
-      "required_pages_per_second_upper_estimate": 0.001728570292832653
+      "required_token_positions_per_second": 28402.479185195734,
+      "required_pages_per_second_upper_estimate": 0.0017295619660867214
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 1546694.2737633225,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28386.19414915223,
-      "required_pages_per_second_upper_estimate": 0.00218524780361636
+      "required_token_positions_per_second": 28402.479185195734,
+      "required_pages_per_second_upper_estimate": 0.002186501470770854
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 22,
-      "required_token_positions_per_second": 28386.19414915223,
-      "required_pages_per_second_upper_estimate": 0.35246152885198395
+      "required_token_positions_per_second": 28402.479185195734,
+      "required_pages_per_second_upper_estimate": 0.3526637344971347
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,14 +158,14 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.038900387710938844,
-    "producer_p95_seconds": 0.13927289700041,
-    "gpu_update_seconds": 0.9739946064881587,
-    "producer_batches_per_second": 25.70668465905286,
-    "gpu_batches_per_second": 1.026699730510425,
-    "producer_headroom": 25.038172208609378,
+    "producer_batch_seconds": 0.04045026681261277,
+    "producer_p95_seconds": 0.13801025099928665,
+    "gpu_update_seconds": 0.9734361504051733,
+    "producer_batches_per_second": 24.721715795659247,
+    "gpu_batches_per_second": 1.0272887436775078,
+    "producer_headroom": 24.0650118555373,
     "sustainable": true,
-    "buffer_seconds": 60.38766560226584,
+    "buffer_seconds": 60.35304132512074,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 68,

metrics/void-byte-v1/1790485060735632541-7579de36/train.jsonl

@@ -10,3 +10,4 @@
 {"step": 86000, "loss": 0.5380822457373142, "lr": 0.0003, "grad_norm": 0.1317840326273743, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12585673405, "tokens_per_second": 149026.47053418861, "data_wait_s": 7.379001544904895e-06, "gpu_peak_gb": 85.46991729736328}
 {"step": 86100, "loss": 0.6486919410526752, "lr": 0.0003, "grad_norm": 0.1274605539704474, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12600191174, "tokens_per_second": 148878.05237954404, "data_wait_s": 1.3258000763016753e-05, "gpu_peak_gb": 85.46991729736328}
 {"step": 86200, "loss": 0.6316159069538116, "lr": 0.0003, "grad_norm": 0.1343899664547231, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12614745751, "tokens_per_second": 149288.384666239, "data_wait_s": 9.057999704964459e-06, "gpu_peak_gb": 85.46991729736328}
+{"step": 86300, "loss": 0.6972736828029156, "lr": 0.0003, "grad_norm": 0.1529532199021092, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12629443000, "tokens_per_second": 150473.8301626082, "data_wait_s": 1.6077001419034787e-05, "gpu_peak_gb": 85.46991729736328}

runs/void-byte-v1/events.out.tfevents.1790485060.0b09bbb2-f535-42d2-bd00-cbae7fb94f51-r2qrq.28895.0.1790485060735632541-7579de36

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c140d403616bc651cc3a8cfc2a8ee69f4e7bc4a8a60b8af252e014cea1828d31
-size 25581
+oid sha256:9ba53cf618b6fb8715581c5b71442fe0fe8d8e18e65a9d31f42861f1a2ccc905
+size 27699