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": 77432.47772706118,
-      "required_pages_per_second_upper_estimate": 0.005356958477259853
+      "required_token_positions_per_second": 77522.40935932283,
+      "required_pages_per_second_upper_estimate": 0.00536318015624985
     },
     {
       "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": 9679.059715882648,
-      "required_pages_per_second_upper_estimate": 0.004052420181289166
+      "required_token_positions_per_second": 9690.301169915354,
+      "required_pages_per_second_upper_estimate": 0.004057126743344406
     },
     {
       "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": 19358.119431765295,
-      "required_pages_per_second_upper_estimate": 0.004031785769697123
+      "required_token_positions_per_second": 19380.602339830708,
+      "required_pages_per_second_upper_estimate": 0.004036468366532991
     },
     {
       "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": 29037.17914764794,
-      "required_pages_per_second_upper_estimate": 0.001740734552696143
+      "required_token_positions_per_second": 29070.903509746062,
+      "required_pages_per_second_upper_estimate": 0.0017427562767098061
     },
     {
       "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": 29037.17914764794,
-      "required_pages_per_second_upper_estimate": 0.0022575129576765337
+      "required_token_positions_per_second": 29070.903509746062,
+      "required_pages_per_second_upper_estimate": 0.002260134878491871
     },
     {
       "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": 29037.17914764794,
-      "required_pages_per_second_upper_estimate": 0.37245775641215406
+      "required_token_positions_per_second": 29070.903509746062,
+      "required_pages_per_second_upper_estimate": 0.3728903363187499
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.04755860790625199,
-    "producer_p95_seconds": 0.11938068800009205,
-    "gpu_update_seconds": 0.9521586053320036,
-    "producer_batches_per_second": 21.026687786387903,
-    "gpu_batches_per_second": 1.0502451948657385,
-    "producer_headroom": 20.02074171743858,
+    "producer_batch_seconds": 0.0490063238906302,
+    "producer_p95_seconds": 0.11353050499997153,
+    "gpu_update_seconds": 0.9510540321091489,
+    "producer_batches_per_second": 20.40552974819635,
+    "gpu_batches_per_second": 1.0514649706939403,
+    "producer_headroom": 19.406761344345323,
     "sustainable": true,
-    "buffer_seconds": 60.93815074124823,
+    "buffer_seconds": 60.867458054985526,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 48,
+    "upload_pages_per_commit": 52,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791033297667653750-ba4975eb/train.jsonl

@@ -2,3 +2,4 @@
 {"step": 600200, "loss": 0.5463886857032776, "lr": 0.0003, "grad_norm": 0.16265245312703575, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 87851654044, "tokens_per_second": 153987.49765921294, "data_wait_s": 1.036000003296067e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 600300, "loss": 0.5745486617088318, "lr": 0.0003, "grad_norm": 0.16983633367320333, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 87866176121, "tokens_per_second": 152576.3584327869, "data_wait_s": 8.928999932322768e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 600400, "loss": 0.5704043507575989, "lr": 0.0003, "grad_norm": 0.16691128189004883, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 87880667765, "tokens_per_second": 152431.60604831862, "data_wait_s": 1.5606000033585588e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 600500, "loss": 0.5799246430397034, "lr": 0.0003, "grad_norm": 0.18639385467417846, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 87895218915, "tokens_per_second": 153047.72347827797, "data_wait_s": 1.4684999996461556e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791033297.83ac8725-d818-41af-bbb1-78dcd88792cf-7zwhs.737.0.1791033297667653750-ba4975eb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:971c7b5dcc545faef68e3a7814dff8938a0a9bf5c3b44dd50467caa6c143d605
-size 8637
+oid sha256:72e4d8dd554638f915a2e5ceb171e530de9f7b44967d7a68a4124a8cd0e15a98
+size 10755