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": 75034.26560634059,
-      "required_pages_per_second_upper_estimate": 0.006503779970150066
+      "required_token_positions_per_second": 75057.26276596382,
+      "required_pages_per_second_upper_estimate": 0.006505773305660439
     },
     {
       "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": 9379.283200792574,
-      "required_pages_per_second_upper_estimate": 0.0037415363015767407
+      "required_token_positions_per_second": 9382.157845745478,
+      "required_pages_per_second_upper_estimate": 0.0037426830404282264
     },
     {
       "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": 18758.566401585147,
-      "required_pages_per_second_upper_estimate": 0.004339464197513719
+      "required_token_positions_per_second": 18764.315691490956,
+      "required_pages_per_second_upper_estimate": 0.004340794194549382
     },
     {
       "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": 28137.849602377722,
-      "required_pages_per_second_upper_estimate": 0.001713447412897224
+      "required_token_positions_per_second": 28146.47353723643,
+      "required_pages_per_second_upper_estimate": 0.0017139725652838297
     },
     {
       "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": 28137.849602377722,
-      "required_pages_per_second_upper_estimate": 0.002166129552943954
+      "required_token_positions_per_second": 28146.47353723643,
+      "required_pages_per_second_upper_estimate": 0.0021667934473219566
     },
     {
       "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": 28137.849602377722,
-      "required_pages_per_second_upper_estimate": 0.3493779207367759
+      "required_token_positions_per_second": 28146.47353723643,
+      "required_pages_per_second_upper_estimate": 0.34948500114526776
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,33 +158,27 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.2575375195392553,
+    "producer_batch_seconds": 0.12556662605476276,
     "producer_p95_seconds": 0.42661202199815307,
-    "gpu_update_seconds": 0.9825910789452678,
-    "producer_batches_per_second": 3.8829293758402237,
-    "gpu_batches_per_second": 1.0177173611971109,
-    "producer_headroom": 3.8153317648751206,
+    "gpu_update_seconds": 0.9822900180878086,
+    "producer_batches_per_second": 7.963899576021696,
+    "gpu_batches_per_second": 1.018029280137313,
+    "producer_headroom": 7.8228590585798425,
     "sustainable": true,
-    "buffer_seconds": 60.9206468946066,
+    "buffer_seconds": 60.90198112144413,
     "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": 482637.801677831,
-    "upload_bytes_per_second": 5867902.487125478,
+    "generation_bytes_per_second": 483536.875749722,
+    "upload_bytes_per_second": 5773531.969764306,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 103.56283700705717,
+    "estimated_full_shards_per_hour": 103.75575737351174,
     "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": [
-    {
-      "seconds": 1.1027901979978196,
-      "bytes": 11991791,
-      "pages": 37,
-      "shards": 3
-    },
     {
       "seconds": 1.5872819810028886,
       "bytes": 6214149,
@@ -562,6 +556,12 @@
       "bytes": 6760567,
       "pages": 31,
       "shards": 2
+    },
+    {
+      "seconds": 1.5065287670004182,
+      "bytes": 7642751,
+      "pages": 35,
+      "shards": 3
     }
   ],
   "cache_key": {

metrics/void-byte-v1/1790519810984249790-cb9c9513/train.jsonl

@@ -403,3 +403,4 @@
 {"step": 155300, "loss": 0.622527826577425, "lr": 0.0003, "grad_norm": 0.16087515427852186, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22728190005, "tokens_per_second": 148964.77667606197, "data_wait_s": 9.969997336156666e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 155400, "loss": 0.619165938347578, "lr": 0.0003, "grad_norm": 0.1476237453690166, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22742793235, "tokens_per_second": 148852.73737805066, "data_wait_s": 1.5402001736219972e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 155500, "loss": 0.5956929791718721, "lr": 0.0003, "grad_norm": 0.1505320523732815, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22757469182, "tokens_per_second": 149098.32961767094, "data_wait_s": 1.2772994523402303e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 155600, "loss": 0.6896321140229702, "lr": 0.0003, "grad_norm": 0.16904787321003917, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22772124076, "tokens_per_second": 149032.15146045198, "data_wait_s": 1.2080003216397017e-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 sha256:098289a34ef409c40febe5edd8b9aa70f1fa660f92c2f2d47d3eb1cc095a15a7
-size 864219
+oid sha256:2018aee8ecfd820d1f136f3e212952e9843e770dccfdd81ee2654877d64db5c3
+size 866337