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": 76199.98851519359,
-      "required_pages_per_second_upper_estimate": 0.006604821877391742
+      "required_token_positions_per_second": 76229.27411781656,
+      "required_pages_per_second_upper_estimate": 0.00660736027920342
     },
     {
       "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": 9524.998564399199,
-      "required_pages_per_second_upper_estimate": 0.003799664338758257
+      "required_token_positions_per_second": 9528.65926472707,
+      "required_pages_per_second_upper_estimate": 0.0038011246468513922
     },
     {
       "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": 19049.997128798397,
-      "required_pages_per_second_upper_estimate": 0.004406881567248884
+      "required_token_positions_per_second": 19057.31852945414,
+      "required_pages_per_second_upper_estimate": 0.004408575244438339
     },
     {
       "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": 28574.995693197594,
-      "required_pages_per_second_upper_estimate": 0.0017400673162998688
+      "required_token_positions_per_second": 28585.97779418121,
+      "required_pages_per_second_upper_estimate": 0.0017407360686311402
     },
     {
       "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": 28574.995693197594,
-      "required_pages_per_second_upper_estimate": 0.0021997822691130963
+      "required_token_positions_per_second": 28585.97779418121,
+      "required_pages_per_second_upper_estimate": 0.0022006277016472166
     },
     {
       "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": 28574.995693197594,
-      "required_pages_per_second_upper_estimate": 0.35480581215090695
+      "required_token_positions_per_second": 28585.97779418121,
+      "required_pages_per_second_upper_estimate": 0.35494217309039583
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,14 +158,14 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.04982462566404777,
-    "producer_p95_seconds": 0.13434242899984383,
-    "gpu_update_seconds": 0.9675592009479018,
-    "producer_batches_per_second": 20.07039664969476,
-    "gpu_batches_per_second": 1.033528490060677,
-    "producer_headroom": 19.41929694508611,
+    "producer_batch_seconds": 0.054874912054685865,
+    "producer_p95_seconds": 0.1411703599997054,
+    "gpu_update_seconds": 0.9671874860837492,
+    "producer_batches_per_second": 18.22326382962482,
+    "gpu_batches_per_second": 1.033925701467781,
+    "producer_headroom": 17.625312731615747,
     "sustainable": true,
-    "buffer_seconds": 60.95622965971781,
+    "buffer_seconds": 60.932811623276194,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 78,

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

@@ -27,3 +27,4 @@
 {"step": 157700, "loss": 0.5980375073850155, "lr": 0.0003, "grad_norm": 0.1388002859641002, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 23080472965, "tokens_per_second": 150772.9096988358, "data_wait_s": 1.421999968442833e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 157800, "loss": 0.602766864001751, "lr": 0.0003, "grad_norm": 0.13362723631035256, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 23094887446, "tokens_per_second": 148526.46023734435, "data_wait_s": 1.6574999790464062e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 157900, "loss": 0.6427027005702257, "lr": 0.0003, "grad_norm": 0.16760281105504934, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 23109366651, "tokens_per_second": 149030.92749643698, "data_wait_s": 1.3122999916959088e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 158000, "loss": 0.5854500038549304, "lr": 0.0003, "grad_norm": 0.13975316095131562, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 23124097242, "tokens_per_second": 149803.65281622653, "data_wait_s": 1.4758000361325685e-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:e94e774d3faedfd0b3908af2f07f6b21c72be9c36fa6457c4c5ebebdee4b3b57
-size 61587
+oid sha256:84ea54f5cc68e137a28cd9973c3dc8d42e92deb894145a5dfd6e3d874b2795db
+size 63705