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": 75047.49851322017,
-      "required_pages_per_second_upper_estimate": 0.006504926964979891
+      "required_token_positions_per_second": 75346.75270619031,
+      "required_pages_per_second_upper_estimate": 0.006530865559973726
     },
     {
       "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": 9380.937314152521,
-      "required_pages_per_second_upper_estimate": 0.0037421961521272224
+      "required_token_positions_per_second": 9418.344088273789,
+      "required_pages_per_second_upper_estimate": 0.003757118273605309
     },
     {
       "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": 18761.874628305042,
-      "required_pages_per_second_upper_estimate": 0.0043402294975425654
+      "required_token_positions_per_second": 18836.688176547577,
+      "required_pages_per_second_upper_estimate": 0.004357536295254997
     },
     {
       "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": 28142.811942457563,
-      "required_pages_per_second_upper_estimate": 0.0017137495933727
+      "required_token_positions_per_second": 28255.032264821366,
+      "required_pages_per_second_upper_estimate": 0.001720583222230127
     },
     {
       "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": 28142.811942457563,
-      "required_pages_per_second_upper_estimate": 0.002166511567619936
+      "required_token_positions_per_second": 28255.032264821366,
+      "required_pages_per_second_upper_estimate": 0.0021751505986812364
     },
     {
       "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": 28142.811942457563,
-      "required_pages_per_second_upper_estimate": 0.34943953639268366
+      "required_token_positions_per_second": 28255.032264821366,
+      "required_pages_per_second_upper_estimate": 0.3508329372191833
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,24 +158,24 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.07890436042183069,
-    "producer_p95_seconds": 0.19544624099944485,
-    "gpu_update_seconds": 0.9824178215215564,
-    "producer_batches_per_second": 12.673570822371525,
-    "gpu_batches_per_second": 1.0178968439835634,
-    "producer_headroom": 12.450741838213395,
+    "producer_batch_seconds": 0.1102363699687885,
+    "producer_p95_seconds": 0.21926446399993438,
+    "gpu_update_seconds": 0.9785159592411032,
+    "producer_batches_per_second": 9.071416269268777,
+    "gpu_batches_per_second": 1.0219557387449858,
+    "producer_headroom": 8.876525592398888,
     "sustainable": true,
-    "buffer_seconds": 60.909904934336495,
+    "buffer_seconds": 60.6679894729484,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 83,
+    "upload_pages_per_commit": 85,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 446436.5591577297,
+    "generation_bytes_per_second": 427876.060916247,
     "upload_bytes_per_second": 6024952.071976909,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 95.7948930840389,
+    "estimated_full_shards_per_hour": 91.8122422277027,
     "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
   },
   "upload_measurements": [

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

@@ -114,3 +114,4 @@
 {"step": 51400, "loss": 0.6452398039400578, "lr": 0.0003, "grad_norm": 0.14211460961088088, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 7522640579, "tokens_per_second": 151092.2827554876, "data_wait_s": 8.70500116434414e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 51500, "loss": 0.6290330532938242, "lr": 0.0003, "grad_norm": 0.13108455622104725, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 7537422050, "tokens_per_second": 151110.25052112326, "data_wait_s": 7.225999070215039e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 51600, "loss": 0.6494944877922535, "lr": 0.0003, "grad_norm": 0.13493877725330178, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 7552234544, "tokens_per_second": 151325.46088628046, "data_wait_s": 1.1256999641773291e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 51700, "loss": 0.6375000514090061, "lr": 0.0003, "grad_norm": 0.13611204968969093, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 7567016266, "tokens_per_second": 150654.87974641632, "data_wait_s": 1.2184998922748491e-05, "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:d956737839e18fd719224caed6be3eb17e8b56c7cd15ae2f3d4b614c6f582772
-size 247419
+oid sha256:75267e423c35bf175576d4c34bda46dfe01d0610b0046427334896c9b683798a
+size 249537