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": 76301.52223210933,
-      "required_pages_per_second_upper_estimate": 0.006613622562639146
+      "required_token_positions_per_second": 76313.46431551958,
+      "required_pages_per_second_upper_estimate": 0.0066146576721622125
     },
     {
       "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": 9537.690279013666,
-      "required_pages_per_second_upper_estimate": 0.003804727253476012
+      "required_token_positions_per_second": 9539.183039439948,
+      "required_pages_per_second_upper_estimate": 0.0038053227379288125
     },
     {
       "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": 19075.380558027333,
-      "required_pages_per_second_upper_estimate": 0.004412753576867901
+      "required_token_positions_per_second": 19078.366078879895,
+      "required_pages_per_second_upper_estimate": 0.004413444224573769
     },
     {
       "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": 28613.070837041,
-      "required_pages_per_second_upper_estimate": 0.001742385892795091
+      "required_token_positions_per_second": 28617.54911831984,
+      "required_pages_per_second_upper_estimate": 0.0017426585966292465
     },
     {
       "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": 28613.070837041,
-      "required_pages_per_second_upper_estimate": 0.002202713399084906
+      "required_token_positions_per_second": 28617.54911831984,
+      "required_pages_per_second_upper_estimate": 0.002203058149574427
     },
     {
       "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": 28613.070837041,
-      "required_pages_per_second_upper_estimate": 0.3552785780081329
+      "required_token_positions_per_second": 28617.54911831984,
+      "required_pages_per_second_upper_estimate": 0.3553341832737728
     }
   ],
   "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": 688946,
-    "producer_batch_seconds": 0.057083336726549305,
-    "producer_p95_seconds": 0.14163955799995165,
-    "gpu_update_seconds": 0.9662716790330779,
-    "producer_batches_per_second": 17.51824713384182,
-    "gpu_batches_per_second": 1.0349056292332537,
-    "producer_headroom": 16.927386071733743,
+    "producer_batch_seconds": 0.051223755554680395,
+    "producer_p95_seconds": 0.13963601000023118,
+    "gpu_update_seconds": 0.9661204698448767,
+    "producer_batches_per_second": 19.522192177660983,
+    "gpu_batches_per_second": 1.035067604105897,
+    "producer_headroom": 18.860789479083806,
     "sustainable": true,
-    "buffer_seconds": 60.875115779083906,
+    "buffer_seconds": 60.86558960022723,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 78,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790434090210070261-a05fc3fb/train.jsonl

@@ -19,3 +19,4 @@
 {"step": 41900, "loss": 0.6494179293513298, "lr": 0.0003, "grad_norm": 0.13507546991520308, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6131418623, "tokens_per_second": 150099.77348306376, "data_wait_s": 1.2029000117763644e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 42000, "loss": 0.6386174354702234, "lr": 0.0003, "grad_norm": 0.13100634088739688, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6146009357, "tokens_per_second": 150437.83926237703, "data_wait_s": 9.286000022257213e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 42100, "loss": 0.6264883726835251, "lr": 0.0003, "grad_norm": 0.13139643423860833, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6160514075, "tokens_per_second": 148933.64423017573, "data_wait_s": 8.6850000116101e-06, "gpu_peak_gb": 85.46427917480469}
+{"step": 42200, "loss": 0.6160119939595461, "lr": 0.0003, "grad_norm": 0.12473146730710874, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6175109717, "tokens_per_second": 150601.47482535624, "data_wait_s": 1.2016999789921101e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790434090.372bd512-57cc-4928-b44a-3b375a0fa071-5spg5.451.0.1790434090210070261-a05fc3fb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:11fa6e45336b209a5ee56948a780d66db021935b4661aa3d40aaf425467081ea
-size 44643
+oid sha256:6c654da748fb31c9ff16ce8ecc7bca6a28ae78a8e82dade1c7a5b49e7bc89f05
+size 46761