Update TensorBoard training metrics

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 598315.063954628,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 79257.51310602436,
-      "required_pages_per_second_upper_estimate": 0.005483218659442028
+      "required_token_positions_per_second": 78832.32975777458,
+      "required_pages_per_second_upper_estimate": 0.005453803488848819
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 8533.957213384612,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9907.189138253045,
-      "required_pages_per_second_upper_estimate": 0.004147933206551593
+      "required_token_positions_per_second": 9854.041219721823,
+      "required_pages_per_second_upper_estimate": 0.004125681282917315
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 380144.35761751566,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19814.37827650609,
-      "required_pages_per_second_upper_estimate": 0.004126812454701754
+      "required_token_positions_per_second": 19708.082439443646,
+      "required_pages_per_second_upper_estimate": 0.004104673835051378
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 43742.16556775435,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29721.567414759134,
-      "required_pages_per_second_upper_estimate": 0.0017817625843090828
+      "required_token_positions_per_second": 29562.12365916547,
+      "required_pages_per_second_upper_estimate": 0.0017722041746177609
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 22018.86276523721,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29721.567414759134,
-      "required_pages_per_second_upper_estimate": 0.0023107211351385813
+      "required_token_positions_per_second": 29562.12365916547,
+      "required_pages_per_second_upper_estimate": 0.0022983250844601367
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 910315.184299418,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 24,
-      "required_token_positions_per_second": 29721.567414759134,
-      "required_pages_per_second_upper_estimate": 0.38123635426378744
+      "required_token_positions_per_second": 29562.12365916547,
+      "required_pages_per_second_upper_estimate": 0.37919118096439847
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,27 +158,40 @@
     "prefetch_batches": 65,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 2.558590814491007,
-    "producer_p95_seconds": 2.558590814491007,
-    "gpu_update_seconds": 0.9302335780000135,
-    "producer_batches_per_second": 0.39084014307263704,
-    "gpu_batches_per_second": 1.0749988214250266,
-    "producer_headroom": 0.3635726247164963,
-    "sustainable": false,
-    "buffer_seconds": 60.46518257000088,
-    "buffer_drain_seconds": 95.0071994358602,
+    "producer_batch_seconds": 0.053032901969219476,
+    "producer_p95_seconds": 0.13808289299993248,
+    "gpu_update_seconds": 0.9352508067000115,
+    "producer_batches_per_second": 18.856218740969602,
+    "gpu_batches_per_second": 1.0692319031816213,
+    "producer_headroom": 17.635293788803697,
+    "sustainable": true,
+    "buffer_seconds": 60.79130243550075,
+    "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 9,
-    "max_pending_pages": 2138,
-    "generation_bytes_per_second": 0,
-    "upload_bytes_per_second": null,
-    "upload_sustainable": null,
+    "upload_pages_per_commit": 18,
+    "max_pending_pages": 4096,
+    "generation_bytes_per_second": 414846.1752693891,
+    "upload_bytes_per_second": 82964.0328963144,
+    "upload_sustainable": false,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 0.0,
+    "estimated_full_shards_per_hour": 89.01633208810095,
     "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": [],
+  "upload_measurements": [
+    {
+      "seconds": 0.9708695179999722,
+      "bytes": 43075,
+      "pages": 1,
+      "shards": 1
+    },
+    {
+      "seconds": 0.8249086619999844,
+      "bytes": 105910,
+      "pages": 3,
+      "shards": 1
+    }
+  ],
   "cache_key": {
     "version": 1,
     "kind": "data",

metrics/void-byte-v1/1790824801708454555-41033672/train.jsonl

@@ -0,0 +1 @@
+{"step": 397200, "loss": 0.6217398047447205, "lr": 0.0003, "grad_norm": 0.1379693229902149, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 58139220888, "tokens_per_second": 152102.9842612424, "data_wait_s": 1.3022000075579854e-05, "gpu_peak_gb": 89.37861251831055}

runs/void-byte-v1/events.out.tfevents.1790824801.afc92292-ba09-4e51-8d82-bc1983323308-sdcdh.469.0.1790824801708454555-41033672

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:24f5c75d5a7fd13da224df91bf8ca10f057a9d45bb13ecc04c67f1bbf3650a54
-size 165
+oid sha256:8b7095a8168f8b38c79a67e97033c32697b08d6628d1ea7ac5ed1071cd41e579
+size 2283