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": 76298.10938336971,
-      "required_pages_per_second_upper_estimate": 0.006613326745560181
+      "required_token_positions_per_second": 76190.95873443011,
+      "required_pages_per_second_upper_estimate": 0.006604039198880934
     },
     {
       "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": 9537.263672921214,
-      "required_pages_per_second_upper_estimate": 0.003804557073927403
+      "required_token_positions_per_second": 9523.869841803764,
+      "required_pages_per_second_upper_estimate": 0.0037992140744390313
     },
     {
       "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": 19074.527345842427,
-      "required_pages_per_second_upper_estimate": 0.004412556201244945
+      "required_token_positions_per_second": 19047.739683607528,
+      "required_pages_per_second_upper_estimate": 0.004406359347033654
     },
     {
       "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": 28611.791018763644,
-      "required_pages_per_second_upper_estimate": 0.0017423079585766877
+      "required_token_positions_per_second": 28571.60952541129,
+      "required_pages_per_second_upper_estimate": 0.0017398611164475323
     },
     {
       "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": 28611.791018763644,
-      "required_pages_per_second_upper_estimate": 0.0022026148751311514
+      "required_token_positions_per_second": 28571.60952541129,
+      "required_pages_per_second_upper_estimate": 0.0021995215925433944
     },
     {
       "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": 28611.791018763644,
-      "required_pages_per_second_upper_estimate": 0.3552626869484044
+      "required_token_positions_per_second": 28571.60952541129,
+      "required_pages_per_second_upper_estimate": 0.35476376727977565
     }
   ],
   "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": 0.05209956435156471,
-    "producer_p95_seconds": 0.133577422000144,
-    "gpu_update_seconds": 0.9663149007997581,
-    "producer_batches_per_second": 19.194018461499226,
-    "gpu_batches_per_second": 1.0348593395096803,
-    "producer_headroom": 18.54746604557235,
+    "producer_batch_seconds": 0.05203348372657146,
+    "producer_p95_seconds": 0.13871924800014312,
+    "gpu_update_seconds": 0.9676738713445652,
+    "producer_batches_per_second": 19.218394164320372,
+    "gpu_batches_per_second": 1.033406015820721,
+    "producer_headroom": 18.597137882013694,
     "sustainable": true,
-    "buffer_seconds": 60.87783875038476,
+    "buffer_seconds": 60.963453894707605,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 77,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -17,3 +17,4 @@
 {"step": 41700, "loss": 0.645052120089531, "lr": 0.0003, "grad_norm": 0.13222859382635524, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6102058721, "tokens_per_second": 151438.91375842685, "data_wait_s": 9.574000159773277e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 41800, "loss": 0.6425942555069923, "lr": 0.0003, "grad_norm": 0.13991135484243736, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6116836242, "tokens_per_second": 152251.9526339524, "data_wait_s": 1.2988999969820725e-05, "gpu_peak_gb": 85.46427917480469}
 {"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}

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:e2fb4ceead31390dae5be4f5c2d1c79c0767d30a7b393aed6e89579d95fae391
-size 40407
+oid sha256:be3aa9b4ac25ace51a7be16b88f6c26da7531165ba0a3297b70dd2468ee9520d
+size 42525