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": 76360.14041154199,
-      "required_pages_per_second_upper_estimate": 0.006618703437865951
+      "required_token_positions_per_second": 76377.82237376756,
+      "required_pages_per_second_upper_estimate": 0.00662023606553714
     },
     {
       "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": 9545.017551442748,
-      "required_pages_per_second_upper_estimate": 0.003807650212000458
+      "required_token_positions_per_second": 9547.227796720945,
+      "required_pages_per_second_upper_estimate": 0.0038085319118880425
     },
     {
       "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": 19090.035102885497,
-      "required_pages_per_second_upper_estimate": 0.0044161436479096565
+      "required_token_positions_per_second": 19094.45559344189,
+      "required_pages_per_second_upper_estimate": 0.004417166250601901
     },
     {
       "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": 28635.052654328243,
-      "required_pages_per_second_upper_estimate": 0.0017437244701382021
+      "required_token_positions_per_second": 28641.683390162838,
+      "required_pages_per_second_upper_estimate": 0.0017441282471617465
     },
     {
       "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": 28635.052654328243,
-      "required_pages_per_second_upper_estimate": 0.0022044056202292425
+      "required_token_positions_per_second": 28641.683390162838,
+      "required_pages_per_second_upper_estimate": 0.002204916072628841
     },
     {
       "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": 28635.052654328243,
-      "required_pages_per_second_upper_estimate": 0.35555151861043055
+      "required_token_positions_per_second": 28641.683390162838,
+      "required_pages_per_second_upper_estimate": 0.3556338501578509
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.04678823423437528,
-    "producer_p95_seconds": 0.11865893499998492,
-    "gpu_update_seconds": 0.9655299165591354,
-    "producer_batches_per_second": 21.37289462540351,
-    "gpu_batches_per_second": 1.0357006891756455,
-    "producer_headroom": 20.636169164293047,
+    "producer_batch_seconds": 0.04835409394531354,
+    "producer_p95_seconds": 0.12721536100002595,
+    "gpu_update_seconds": 0.9653063901088955,
+    "producer_batches_per_second": 20.680772162352135,
+    "gpu_batches_per_second": 1.0359405161372555,
+    "producer_headroom": 19.963281520704676,
     "sustainable": true,
-    "buffer_seconds": 60.828384743225534,
+    "buffer_seconds": 60.81430257686042,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 60,
+    "upload_pages_per_commit": 64,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -6,3 +6,4 @@
 {"step": 115600, "loss": 0.7220963202416897, "lr": 0.0003, "grad_norm": 0.16984515292624902, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16916387672, "tokens_per_second": 151576.28877827647, "data_wait_s": 1.5824000001884997e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 115700, "loss": 0.6949176611378789, "lr": 0.0003, "grad_norm": 0.1675070391333178, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16931012208, "tokens_per_second": 151135.76577111467, "data_wait_s": 1.2496000181272393e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 115800, "loss": 0.603627547621727, "lr": 0.0003, "grad_norm": 0.15810096959168768, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16945791532, "tokens_per_second": 152467.27985714795, "data_wait_s": 9.756999816090683e-06, "gpu_peak_gb": 85.46427917480469}
+{"step": 115900, "loss": 0.5959490407258272, "lr": 0.0003, "grad_norm": 0.14769660827695635, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16960422649, "tokens_per_second": 151080.4202959854, "data_wait_s": 1.5132000044104643e-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:fe9590086e8b5c110c5e1e636f6a8239fedb2da224c15550537f48ad95d46038
-size 17109
+oid sha256:8be8bf4481060171308d51278563ef3ffe437d992f4615b84fb4663ca78e8a33
+size 19227