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": 76019.39367997316,
-      "required_pages_per_second_upper_estimate": 0.006589168374788261
+      "required_token_positions_per_second": 75082.0941356737,
+      "required_pages_per_second_upper_estimate": 0.006507925625852361
     },
     {
       "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": 9502.424209996645,
-      "required_pages_per_second_upper_estimate": 0.0037906590912704023
+      "required_token_positions_per_second": 9385.261766959213,
+      "required_pages_per_second_upper_estimate": 0.003743921241008143
     },
     {
       "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": 19004.84841999329,
-      "required_pages_per_second_upper_estimate": 0.004396437208056958
+      "required_token_positions_per_second": 18770.523533918426,
+      "required_pages_per_second_upper_estimate": 0.004342230269640682
     },
     {
       "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": 28507.272629989937,
-      "required_pages_per_second_upper_estimate": 0.001735943337065975
+      "required_token_positions_per_second": 28155.785300877636,
+      "required_pages_per_second_upper_estimate": 0.0017145396028345312
     },
     {
       "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": 28507.272629989937,
-      "required_pages_per_second_upper_estimate": 0.0021945687602379334
+      "required_token_positions_per_second": 28155.785300877636,
+      "required_pages_per_second_upper_estimate": 0.002167510292663668
     },
     {
       "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": 28507.272629989937,
-      "required_pages_per_second_upper_estimate": 0.35396491836038013
+      "required_token_positions_per_second": 28155.785300877636,
+      "required_pages_per_second_upper_estimate": 0.3496006220852234
     }
   ],
   "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": 0.12020363978908222,
-    "producer_p95_seconds": 0.32443845699890517,
-    "gpu_update_seconds": 0.9698577748512507,
-    "producer_batches_per_second": 8.319215639016177,
-    "gpu_batches_per_second": 1.0310790158416498,
-    "producer_headroom": 8.068455968163954,
+    "producer_batch_seconds": 0.3503327449921301,
+    "producer_p95_seconds": 0.32461810300083016,
+    "gpu_update_seconds": 0.981965152260846,
+    "producer_batches_per_second": 2.854429151412792,
+    "gpu_batches_per_second": 1.0183660771440117,
+    "producer_headroom": 2.80294995628486,
     "sustainable": true,
-    "buffer_seconds": 60.131182040777546,
+    "buffer_seconds": 60.881839440172456,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 78,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 396540.89214918984,
+    "generation_bytes_per_second": 354908.08680699335,
     "upload_bytes_per_second": 3421513.6051928457,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 85.0884444556882,
+    "estimated_full_shards_per_hour": 76.15501359135962,
     "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

@@ -268,3 +268,4 @@
 {"step": 66800, "loss": 0.6008422300219536, "lr": 0.0003, "grad_norm": 0.1434360788227507, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9776487304, "tokens_per_second": 148791.11147907007, "data_wait_s": 1.032299769576639e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 66900, "loss": 0.608786091208458, "lr": 0.0003, "grad_norm": 0.14827182531631294, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9791009949, "tokens_per_second": 148319.05762522173, "data_wait_s": 8.881001122063026e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 67000, "loss": 0.6176154986023903, "lr": 0.0003, "grad_norm": 0.18369370092204232, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9805506913, "tokens_per_second": 147777.17644516315, "data_wait_s": 1.4841003576293588e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 67100, "loss": 0.6024973634630442, "lr": 0.0003, "grad_norm": 0.1236716504339559, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 9820152581, "tokens_per_second": 149987.07305735402, "data_wait_s": 7.930997526273131e-06, "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 sha256:0f9eb5b47167c6b8946430f04f7a27820040b6d80ef27ab13dcdb07cf2ad80f2
-size 575940
+oid sha256:423177a8eb6ecffeebe8786885d1c43c5da7954d44c2d66fd5398e5f340f3cb8
+size 578058