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": 74755.21274394525,
-      "required_pages_per_second_upper_estimate": 0.006479592375290651
+      "required_token_positions_per_second": 74668.13011026407,
+      "required_pages_per_second_upper_estimate": 0.006472044273312084
     },
     {
       "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": 9344.401592993156,
-      "required_pages_per_second_upper_estimate": 0.0037276215066990408
+      "required_token_positions_per_second": 9333.516263783009,
+      "required_pages_per_second_upper_estimate": 0.0037232791861269384
     },
     {
       "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": 18688.80318598631,
-      "required_pages_per_second_upper_estimate": 0.004323325705375589
+      "required_token_positions_per_second": 18667.032527566018,
+      "required_pages_per_second_upper_estimate": 0.004318289446700556
     },
     {
       "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": 28033.20477897947,
-      "required_pages_per_second_upper_estimate": 0.0017070750921812277
+      "required_token_positions_per_second": 28000.548791349025,
+      "required_pages_per_second_upper_estimate": 0.0017050865138672593
     },
     {
       "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": 28033.20477897947,
-      "required_pages_per_second_upper_estimate": 0.002158073704763345
+      "required_token_positions_per_second": 28000.548791349025,
+      "required_pages_per_second_upper_estimate": 0.002155559756437994
     },
     {
       "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": 21,
-      "required_token_positions_per_second": 28033.20477897947,
-      "required_pages_per_second_upper_estimate": 0.34807858225386434
+      "required_token_positions_per_second": 28000.548791349025,
+      "required_pages_per_second_upper_estimate": 0.3476731041800542
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,24 +158,24 @@
     "prefetch_batches": 61,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.61770679579638,
-    "producer_p95_seconds": 0.4195486729950062,
-    "gpu_update_seconds": 0.9862589817319669,
-    "producer_batches_per_second": 1.6188910447565137,
-    "gpu_batches_per_second": 1.0139324645174865,
-    "producer_headroom": 1.5966458333365592,
+    "producer_batch_seconds": 0.1216597082577664,
+    "producer_p95_seconds": 0.3963894780026749,
+    "gpu_update_seconds": 0.9874092185129618,
+    "producer_batches_per_second": 8.21964818361434,
+    "gpu_batches_per_second": 1.0127513307056217,
+    "producer_headroom": 8.11615638943412,
     "sustainable": true,
-    "buffer_seconds": 60.16179788564998,
+    "buffer_seconds": 60.231962329290674,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 241361.25532943383,
+    "generation_bytes_per_second": 241255.37758607176,
     "upload_bytes_per_second": 5923325.970069969,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 51.790506791231735,
+    "estimated_full_shards_per_hour": 51.76778789221396,
     "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-v1/1790519810984249790-cb9c9513/train.jsonl

@@ -388,3 +388,4 @@
 {"step": 153800, "loss": 0.5959294298663735, "lr": 0.0003, "grad_norm": 0.14721382194840682, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22508767734, "tokens_per_second": 151025.75518803636, "data_wait_s": 1.1620999430306256e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 153900, "loss": 0.5766774993389845, "lr": 0.0003, "grad_norm": 0.13650398248971385, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22523255470, "tokens_per_second": 125261.97257758408, "data_wait_s": 1.1034004273824394e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 154000, "loss": 0.5860744910314679, "lr": 0.0003, "grad_norm": 0.1366348513136134, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22537841869, "tokens_per_second": 149215.42948124663, "data_wait_s": 1.0882999049499631e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 154100, "loss": 0.6051191464066505, "lr": 0.0003, "grad_norm": 0.14220059304235802, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22552412707, "tokens_per_second": 149023.50628410318, "data_wait_s": 1.20009935926646e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790519810.cad0b19a-59f2-43f0-8fb4-2a94097843e2-tjh7c.1336.0.1790519810984249790-cb9c9513

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f39d6c7dcd61c579fa78e14da1fd751f4c93b6f5f1e360547926887efca8ffd1
-size 831666
+oid sha256:cb22ed5a3b1bc3310d75c8bde218ec10b35cd6cfae67527a634503e6a62328d3
+size 833784