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": 78334.50306681794,
-      "required_pages_per_second_upper_estimate": 0.005419362683251371
+      "required_token_positions_per_second": 77918.72783002397,
+      "required_pages_per_second_upper_estimate": 0.005390598387638501
     },
     {
       "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": 9791.812883352242,
-      "required_pages_per_second_upper_estimate": 0.004099627577954804
+      "required_token_positions_per_second": 9739.840978752996,
+      "required_pages_per_second_upper_estimate": 0.004077868026795881
     },
     {
       "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": 19583.625766704485,
-      "required_pages_per_second_upper_estimate": 0.004078752792262986
+      "required_token_positions_per_second": 19479.681957505993,
+      "required_pages_per_second_upper_estimate": 0.0040571040379895245
     },
     {
       "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": 29375.43865005673,
-      "required_pages_per_second_upper_estimate": 0.0017610126933731956
+      "required_token_positions_per_second": 29219.522936258985,
+      "required_pages_per_second_upper_estimate": 0.0017516657844005312
     },
     {
       "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": 29375.43865005673,
-      "required_pages_per_second_upper_estimate": 0.0022838111461425067
+      "required_token_positions_per_second": 29219.522936258985,
+      "required_pages_per_second_upper_estimate": 0.002271689385195479
     },
     {
       "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": 23,
-      "required_token_positions_per_second": 29375.43865005673,
-      "required_pages_per_second_upper_estimate": 0.3767965861142973
+      "required_token_positions_per_second": 29219.522936258985,
+      "required_pages_per_second_upper_estimate": 0.3747966667469502
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,18 +158,18 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 1.7186897505000616,
-    "producer_p95_seconds": 3.328129581999974,
-    "gpu_update_seconds": 0.9411944559999483,
-    "producer_batches_per_second": 0.5818385777357693,
-    "gpu_batches_per_second": 1.0624796965442973,
-    "producer_headroom": 0.547623243651801,
-    "sustainable": false,
-    "buffer_seconds": 60.23644518399669,
-    "buffer_drain_seconds": 133.15548232463138,
+    "producer_batch_seconds": 0.049711916124994815,
+    "producer_p95_seconds": 0.1068931170000269,
+    "gpu_update_seconds": 0.9462166805499463,
+    "producer_batches_per_second": 20.115901336122644,
+    "gpu_batches_per_second": 1.0568403839792746,
+    "producer_headroom": 19.034001388536197,
+    "sustainable": true,
+    "buffer_seconds": 60.557867555196566,
+    "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 7,
-    "max_pending_pages": 1746,
+    "upload_pages_per_commit": 18,
+    "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,
     "upload_sustainable": null,

metrics/void-byte-v1/1791300814123050567-72be911d/train.jsonl

@@ -0,0 +1 @@
+{"step": 800100, "loss": 0.5328859686851501, "lr": 3e-05, "grad_norm": 0.2662423412847009, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117111234819, "tokens_per_second": 153883.04856248165, "data_wait_s": 8.258999969257275e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791300814.0c9d201b-2dbc-40fb-b739-4ee2350a0f54-7b2tx.840.0.1791300814123050567-72be911d

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1a6260d564b11ce0370bfcd3d00419c28da2e1ec98ab269e82947a89fd91aae8
-size 165
+oid sha256:c2e6ec513ccaa81fa9b6de55f4a4ee5bebedcf1abd2d60f289d5a6d0c61ed5ef
+size 2283