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": 76149.23009574851,
-      "required_pages_per_second_upper_estimate": 0.0066004222662402814
+      "required_token_positions_per_second": 76170.5370030234,
+      "required_pages_per_second_upper_estimate": 0.006602269094961006
     },
     {
       "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": 9518.653761968564,
-      "required_pages_per_second_upper_estimate": 0.0037971333022054268
+      "required_token_positions_per_second": 9521.317125377926,
+      "required_pages_per_second_upper_estimate": 0.003798195757690253
     },
     {
       "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": 19037.30752393713,
-      "required_pages_per_second_upper_estimate": 0.004403946050493646
+      "required_token_positions_per_second": 19042.63425075585,
+      "required_pages_per_second_upper_estimate": 0.004405178294995968
     },
     {
       "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": 28555.96128590569,
-      "required_pages_per_second_upper_estimate": 0.0017389082207616088
+      "required_token_positions_per_second": 28563.951376133777,
+      "required_pages_per_second_upper_estimate": 0.0017393947753357357
     },
     {
       "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": 28555.96128590569,
-      "required_pages_per_second_upper_estimate": 0.002198316947749154
+      "required_token_positions_per_second": 28563.951376133777,
+      "required_pages_per_second_upper_estimate": 0.0021989320470128973
     },
     {
       "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": 28555.96128590569,
-      "required_pages_per_second_upper_estimate": 0.35456946851640475
+      "required_token_positions_per_second": 28563.951376133777,
+      "required_pages_per_second_upper_estimate": 0.35466867869592583
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.05450127412502326,
-    "producer_p95_seconds": 0.15067984999996042,
-    "gpu_update_seconds": 0.9682041421468858,
-    "producer_batches_per_second": 18.348194901022843,
-    "gpu_batches_per_second": 1.032840034935825,
-    "producer_headroom": 17.764798304088686,
+    "producer_batch_seconds": 0.058898677953145295,
+    "producer_p95_seconds": 0.16043277299968395,
+    "gpu_update_seconds": 0.9679333099236723,
+    "producer_batches_per_second": 16.978309781342013,
+    "gpu_batches_per_second": 1.03312902836132,
+    "producer_headroom": 16.433871583563835,
     "sustainable": true,
-    "buffer_seconds": 60.02865681310692,
+    "buffer_seconds": 60.01186521526768,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 78,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790519810984249790-cb9c9513/train.jsonl

@@ -34,3 +34,4 @@
 {"step": 118400, "loss": 0.6232305392622948, "lr": 0.0003, "grad_norm": 0.14040777355668516, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17326463478, "tokens_per_second": 151948.54826641432, "data_wait_s": 1.3704000139114214e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 118500, "loss": 0.5947024971246719, "lr": 0.0003, "grad_norm": 0.13812125208627182, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17341101633, "tokens_per_second": 150676.36652273164, "data_wait_s": 1.222200035044807e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 118600, "loss": 0.6693190056830645, "lr": 0.0003, "grad_norm": 0.1557173263448192, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17355731774, "tokens_per_second": 150646.1035705667, "data_wait_s": 1.3274000593810342e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 118700, "loss": 0.6312964111566544, "lr": 0.0003, "grad_norm": 0.14389962333335488, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17370383301, "tokens_per_second": 151218.79794761713, "data_wait_s": 1.2882000191893894e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790519810.cad0b19a-59f2-43f0-8fb4-2a94097843e2-tjh7c.1336.0.1790519810984249790-cb9c9513

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:102d1a7466a004c8912c7b8a5d722a90bc2f7460036c6fd9523279378a23c8be
-size 76413
+oid sha256:d12b9e3991182757c2d6f795bc79d43b650b163295d1fa8f38e38bb5890a0bf3
+size 78531