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": 76803.50084677598,
-      "required_pages_per_second_upper_estimate": 0.005313444397254291
+      "required_token_positions_per_second": 76934.11806200187,
+      "required_pages_per_second_upper_estimate": 0.005322480799277313
     },
     {
       "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": 9600.437605846997,
-      "required_pages_per_second_upper_estimate": 0.004019502745633594
+      "required_token_positions_per_second": 9616.764757750234,
+      "required_pages_per_second_upper_estimate": 0.004026338583185777
     },
     {
       "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": 19200.875211693994,
-      "required_pages_per_second_upper_estimate": 0.003999035945465215
+      "required_token_positions_per_second": 19233.529515500468,
+      "required_pages_per_second_upper_estimate": 0.004005836975796202
     },
     {
       "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": 28801.312817540995,
-      "required_pages_per_second_upper_estimate": 0.0017265947263532646
+      "required_token_positions_per_second": 28850.2942732507,
+      "required_pages_per_second_upper_estimate": 0.001729531089832709
     },
     {
       "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": 28801.312817540995,
-      "required_pages_per_second_upper_estimate": 0.0022391753879770623
+      "required_token_positions_per_second": 28850.2942732507,
+      "required_pages_per_second_upper_estimate": 0.002242983480711837
     },
     {
       "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": 28801.312817540995,
-      "required_pages_per_second_upper_estimate": 0.3694323163830761
+      "required_token_positions_per_second": 28850.2942732507,
+      "required_pages_per_second_upper_estimate": 0.37006059790473056
     }
   ],
   "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": 720750,
-    "producer_batch_seconds": 0.055625471226576906,
-    "producer_p95_seconds": 0.14179542500005482,
-    "gpu_update_seconds": 0.9599562414099893,
-    "producer_batches_per_second": 17.97737579474593,
-    "gpu_batches_per_second": 1.0417141499399953,
-    "producer_headroom": 17.257494098339222,
+    "producer_batch_seconds": 0.0569738248906404,
+    "producer_p95_seconds": 0.1358660040000359,
+    "gpu_update_seconds": 0.9583264467993508,
+    "producer_batches_per_second": 17.551919709085197,
+    "gpu_batches_per_second": 1.0434857593044957,
+    "producer_headroom": 16.82046884931511,
     "sustainable": true,
-    "buffer_seconds": 60.47724320882932,
+    "buffer_seconds": 60.3745661483591,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 78,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791209746770022608-f3fe5fb7/train.jsonl

@@ -24,3 +24,4 @@
 {"step": 722400, "loss": 0.5104002952575684, "lr": 5e-05, "grad_norm": 0.2041116847258441, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105739363686, "tokens_per_second": 153919.0252666146, "data_wait_s": 1.1156999789818656e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 722500, "loss": 0.5125147104263306, "lr": 5e-05, "grad_norm": 0.22152461389744427, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105754059450, "tokens_per_second": 152771.07039198958, "data_wait_s": 1.425099981133826e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 722600, "loss": 0.5423128604888916, "lr": 5e-05, "grad_norm": 0.22533629143302208, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105768734296, "tokens_per_second": 152710.74050596353, "data_wait_s": 1.868499975898885e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 722700, "loss": 0.5818638801574707, "lr": 5e-05, "grad_norm": 0.23183046311760666, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105783284226, "tokens_per_second": 151330.08964284003, "data_wait_s": 1.099399969461956e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791209746.ebd20fd8-0561-42d8-bf2b-56d08e7106ef-6qqk2.329.0.1791209746770022608-f3fe5fb7

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cd1a5ae35049315188dd3ec8d810068aaf6e4aecb5071bf5122d70ed981f506e
-size 55233
+oid sha256:bbbc902a6581dbb4ed281533ee092625c546e3a0f61da1effbe9416cdacbff08
+size 57351