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": 75050.91323713194,
-      "required_pages_per_second_upper_estimate": 0.006505222944594025
+      "required_token_positions_per_second": 75377.78131120224,
+      "required_pages_per_second_upper_estimate": 0.006533555040814879
     },
     {
       "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": 9381.364154641493,
-      "required_pages_per_second_upper_estimate": 0.0037423664251801073
+      "required_token_positions_per_second": 9422.22266390028,
+      "required_pages_per_second_upper_estimate": 0.00375866549541259
     },
     {
       "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": 18762.728309282986,
-      "required_pages_per_second_upper_estimate": 0.004340426981612541
+      "required_token_positions_per_second": 18844.44532780056,
+      "required_pages_per_second_upper_estimate": 0.004359330775676175
     },
     {
       "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": 28144.092463924477,
-      "required_pages_per_second_upper_estimate": 0.001713827570411658
+      "required_token_positions_per_second": 28266.66799170084,
+      "required_pages_per_second_upper_estimate": 0.001721291776949146
     },
     {
       "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": 28144.092463924477,
-      "required_pages_per_second_upper_estimate": 0.0021666101457071697
+      "required_token_positions_per_second": 28266.66799170084,
+      "required_pages_per_second_upper_estimate": 0.0021760463491461717
     },
     {
       "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": 28144.092463924477,
-      "required_pages_per_second_upper_estimate": 0.34945543618367303
+      "required_token_positions_per_second": 28266.66799170084,
+      "required_pages_per_second_upper_estimate": 0.3509774140047536
     }
   ],
   "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": 0.09538216467964844,
-    "producer_p95_seconds": 0.2938880830006383,
-    "gpu_update_seconds": 0.9823731227233432,
-    "producer_batches_per_second": 10.484140335445423,
-    "gpu_batches_per_second": 1.017943159140787,
-    "producer_headroom": 10.29933768040128,
+    "producer_batch_seconds": 0.07766549794510524,
+    "producer_p95_seconds": 0.27626811400114093,
+    "gpu_update_seconds": 0.9781131616969329,
+    "producer_batches_per_second": 12.87573023360785,
+    "gpu_batches_per_second": 1.0223765911350131,
+    "producer_headroom": 12.593921207950963,
     "sustainable": true,
-    "buffer_seconds": 60.90713360884728,
+    "buffer_seconds": 60.64301602520984,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 84,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 547035.6375764761,
+    "generation_bytes_per_second": 547265.7875905216,
     "upload_bytes_per_second": 5189441.022970066,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 117.38111348601066,
+    "estimated_full_shards_per_hour": 117.43049832140672,
     "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

@@ -240,3 +240,4 @@
 {"step": 64000, "loss": 0.6302631683647633, "lr": 0.0003, "grad_norm": 0.13780771526704533, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9366397123, "tokens_per_second": 148638.90808606657, "data_wait_s": 1.1954001820413396e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 64100, "loss": 0.5892891921103001, "lr": 0.0003, "grad_norm": 0.1285240648130058, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9380929376, "tokens_per_second": 148404.10053383745, "data_wait_s": 1.1047002772102132e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 64200, "loss": 0.6446213405579329, "lr": 0.0003, "grad_norm": 0.1374295009308241, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9395625453, "tokens_per_second": 150108.32928739092, "data_wait_s": 8.966002496890724e-06, "gpu_peak_gb": 85.46427917480469}
+{"step": 64300, "loss": 0.6748886741697788, "lr": 0.0003, "grad_norm": 0.1366752813543412, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9410240760, "tokens_per_second": 149030.21262801863, "data_wait_s": 1.0944000678136945e-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 sha256:42b4511264f5c738ef55625c6e70247732371b725e8681e288fd0d634d31c8f6
-size 515853
+oid sha256:292162584a1dbea8b682c784a48b18c9d0f258781c5efacacade3273d38919cb
+size 517971