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": 76419.77905295075,
-      "required_pages_per_second_upper_estimate": 0.006623872764150508
+      "required_token_positions_per_second": 76402.94839768221,
+      "required_pages_per_second_upper_estimate": 0.006622413925608737
     },
     {
       "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": 9552.472381618843,
-      "required_pages_per_second_upper_estimate": 0.003810624055217346
+      "required_token_positions_per_second": 9550.368549710276,
+      "required_pages_per_second_upper_estimate": 0.0038097848052139286
     },
     {
       "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": 19104.944763237687,
-      "required_pages_per_second_upper_estimate": 0.004419592735431075
+      "required_token_positions_per_second": 19100.737099420552,
+      "required_pages_per_second_upper_estimate": 0.0044186193664593355
     },
     {
       "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": 28657.41714485653,
-      "required_pages_per_second_upper_estimate": 0.0017450863502739612
+      "required_token_positions_per_second": 28651.105649130826,
+      "required_pages_per_second_upper_estimate": 0.0017447020132981249
     },
     {
       "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": 28657.41714485653,
-      "required_pages_per_second_upper_estimate": 0.002206127300618982
+      "required_token_positions_per_second": 28651.105649130826,
+      "required_pages_per_second_upper_estimate": 0.0022056414242066793
     },
     {
       "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": 28657.41714485653,
-      "required_pages_per_second_upper_estimate": 0.3558292107336569
+      "required_token_positions_per_second": 28651.105649130826,
+      "required_pages_per_second_upper_estimate": 0.3557508430799611
     }
   ],
   "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": 0.05571386503906428,
-    "producer_p95_seconds": 0.14045609500021783,
-    "gpu_update_seconds": 0.9647764088524042,
-    "producer_batches_per_second": 17.948853472988116,
-    "gpu_batches_per_second": 1.0365095900194057,
-    "producer_headroom": 17.316630396687476,
+    "producer_batch_seconds": 0.04706295378907299,
+    "producer_p95_seconds": 0.12028107400010413,
+    "gpu_update_seconds": 0.9649889375504342,
+    "producer_batches_per_second": 21.248135093300892,
+    "gpu_batches_per_second": 1.0362813096473822,
+    "producer_headroom": 20.504215308612523,
     "sustainable": true,
-    "buffer_seconds": 60.78091375770147,
+    "buffer_seconds": 60.794303065677354,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 44,
+    "upload_pages_per_commit": 51,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790563279934888633-82973089/train.jsonl

@@ -3,3 +3,4 @@
 {"step": 155300, "loss": 0.6224912591278553, "lr": 0.0003, "grad_norm": 0.16059344946226572, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22728190005, "tokens_per_second": 150311.51459487772, "data_wait_s": 1.39969999963796e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 155400, "loss": 0.6191136613488197, "lr": 0.0003, "grad_norm": 0.1460772445792664, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22742793235, "tokens_per_second": 150671.66686189847, "data_wait_s": 1.2129999959142879e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 155500, "loss": 0.5957893170416355, "lr": 0.0003, "grad_norm": 0.15092513045765893, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22757469182, "tokens_per_second": 151500.14020036277, "data_wait_s": 1.4856999996482045e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 155600, "loss": 0.6897633150219917, "lr": 0.0003, "grad_norm": 0.1700488471682717, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22772124076, "tokens_per_second": 150726.57252580667, "data_wait_s": 1.6490000007252092e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790563279.bdfa290c-aca9-4e96-a5d2-8425921a3278-d85hq.1304.0.1790563279934888633-82973089

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:85e29c1daa8cf63ba276ae87a2f2454aee0d300a1c0d8bf88b84f283996d1391
-size 10755
+oid sha256:72ca11aff7326e63c874c2ec65e51b6d010401d7f48d2c8e2e15d9af346db47b
+size 12873