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": 75949.70314674686,
-      "required_pages_per_second_upper_estimate": 0.006583127775997247
+      "required_token_positions_per_second": 75839.6536586954,
+      "required_pages_per_second_upper_estimate": 0.006573588965290824
     },
     {
       "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": 9493.712893343358,
-      "required_pages_per_second_upper_estimate": 0.0037871840168116153
+      "required_token_positions_per_second": 9479.956707336925,
+      "required_pages_per_second_upper_estimate": 0.0037816964685403404
     },
     {
       "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": 18987.425786686716,
-      "required_pages_per_second_upper_estimate": 0.0043924067884693555
+      "required_token_positions_per_second": 18959.91341467385,
+      "required_pages_per_second_upper_estimate": 0.004386042285405389
     },
     {
       "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": 28481.138680030075,
-      "required_pages_per_second_upper_estimate": 0.001734351916627658
+      "required_token_positions_per_second": 28439.870122010776,
+      "required_pages_per_second_upper_estimate": 0.0017318388779636732
     },
     {
       "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": 28481.138680030075,
-      "required_pages_per_second_upper_estimate": 0.0021925568964266196
+      "required_token_positions_per_second": 28439.870122010776,
+      "required_pages_per_second_upper_estimate": 0.0021893799285916137
     },
     {
       "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": 28481.138680030075,
-      "required_pages_per_second_upper_estimate": 0.35364042216658276
+      "required_token_positions_per_second": 28439.870122010776,
+      "required_pages_per_second_upper_estimate": 0.35312800479296197
     }
   ],
   "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": 0.05528960468749489,
-    "producer_p95_seconds": 0.12711867200005145,
-    "gpu_update_seconds": 0.9707477046690468,
-    "producer_batches_per_second": 18.086582561987004,
-    "gpu_batches_per_second": 1.0301337774895136,
-    "producer_headroom": 17.557508507356093,
+    "producer_batch_seconds": 0.05270264038279393,
+    "producer_p95_seconds": 0.15259611099963877,
+    "gpu_update_seconds": 0.9721563383161193,
+    "producer_batches_per_second": 18.974381411191583,
+    "gpu_batches_per_second": 1.0286411357787462,
+    "producer_headroom": 18.44606515451745,
     "sustainable": true,
-    "buffer_seconds": 60.1863576894809,
+    "buffer_seconds": 60.273692975599396,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 66,

metrics/void-byte-v1/1790606918076636751-3952f1c5/train.jsonl

@@ -15,3 +15,4 @@
 {"step": 196500, "loss": 0.5726331695914268, "lr": 0.0003, "grad_norm": 0.13443351313904545, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28759668846, "tokens_per_second": 150724.4681650211, "data_wait_s": 1.4403000022866763e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 196600, "loss": 0.5984502658247948, "lr": 0.0003, "grad_norm": 0.14993225711578184, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28774382558, "tokens_per_second": 150908.8984723203, "data_wait_s": 2.100000028804061e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 196700, "loss": 0.6545633338391781, "lr": 0.0003, "grad_norm": 0.1603858787251228, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28789175919, "tokens_per_second": 151396.29797880666, "data_wait_s": 1.4754999938304536e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 196800, "loss": 0.6597261168062687, "lr": 0.0003, "grad_norm": 0.1597154006396669, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28803846116, "tokens_per_second": 150012.79519459582, "data_wait_s": 1.6419000075984513e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790606918.0a7f8bdc-23d3-4ecb-bd0c-4a4bf460026e-rl2cx.1192.0.1790606918076636751-3952f1c5

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:906aaf190fabdc8c6966b47876116b35b529ddb7eca5d96e9c858589d79a66ad
-size 36171
+oid sha256:a327cfb08f08e3d83acb2ef2aa918468a046587e6bb5f0846ec2d42316e170c3
+size 38289