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": 75920.86785385358,
-      "required_pages_per_second_upper_estimate": 0.006580628405891639
+      "required_token_positions_per_second": 76070.8920845972,
+      "required_pages_per_second_upper_estimate": 0.0065936321259795596
     },
     {
       "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": 9490.108481731697,
-      "required_pages_per_second_upper_estimate": 0.003785746163128968
+      "required_token_positions_per_second": 9508.86151057465,
+      "required_pages_per_second_upper_estimate": 0.0037932270267171894
     },
     {
       "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": 18980.216963463394,
-      "required_pages_per_second_upper_estimate": 0.004390739154087595
+      "required_token_positions_per_second": 19017.7230211493,
+      "required_pages_per_second_upper_estimate": 0.00439941552044916
     },
     {
       "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": 28470.325445195092,
-      "required_pages_per_second_upper_estimate": 0.0017336934473588662
+      "required_token_positions_per_second": 28526.584531723947,
+      "required_pages_per_second_upper_estimate": 0.0017371193305598582
     },
     {
       "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": 28470.325445195092,
-      "required_pages_per_second_upper_estimate": 0.0021917244636760727
+      "required_token_positions_per_second": 28526.584531723947,
+      "required_pages_per_second_upper_estimate": 0.002196055443892183
     },
     {
       "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": 28470.325445195092,
-      "required_pages_per_second_upper_estimate": 0.35350615797950125
+      "required_token_positions_per_second": 28526.584531723947,
+      "required_pages_per_second_upper_estimate": 0.3542047075471392
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.05407117349999169,
-    "producer_p95_seconds": 0.1545213330000479,
-    "gpu_update_seconds": 0.9711164016450021,
-    "producer_batches_per_second": 18.494142724684046,
-    "gpu_batches_per_second": 1.0297426737990123,
-    "producer_headroom": 17.959965334304265,
+    "producer_batch_seconds": 0.05324157288282194,
+    "producer_p95_seconds": 0.15376205499978823,
+    "gpu_update_seconds": 0.969201201400508,
+    "producer_batches_per_second": 18.78231513184021,
+    "gpu_batches_per_second": 1.0317775076578395,
+    "producer_headroom": 18.203842390862473,
     "sustainable": true,
-    "buffer_seconds": 60.20921690199013,
+    "buffer_seconds": 60.090474486831496,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 81,
+    "upload_pages_per_commit": 82,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -24,3 +24,4 @@
 {"step": 157400, "loss": 0.6493245474994183, "lr": 0.0003, "grad_norm": 0.15396559682025537, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 23036282419, "tokens_per_second": 151325.8073817866, "data_wait_s": 1.5146999885473633e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 157500, "loss": 0.5908354483544827, "lr": 0.0003, "grad_norm": 0.13633144832829183, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 23051116604, "tokens_per_second": 152658.71742647485, "data_wait_s": 1.6008999864425277e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 157600, "loss": 0.6076018586754799, "lr": 0.0003, "grad_norm": 0.1378355336076298, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 23065827092, "tokens_per_second": 151324.33062156025, "data_wait_s": 1.2961999800609192e-05, "gpu_peak_gb": 85.46427917480469}
+{"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}

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:b6721a9ce85e30c66a9a70e5526d31684ad2c7a15a2ef00b60d4f7d0d9e7bd70
-size 55233
+oid sha256:ade452d2ac4cba8b8242e715953aac0dc7348d2b1df209f0f139574020668ce3
+size 57351