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": 75833.96945009574,
-      "required_pages_per_second_upper_estimate": 0.0065730962724959005
+      "required_token_positions_per_second": 75924.30636779054,
+      "required_pages_per_second_upper_estimate": 0.006580926447564861
     },
     {
       "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": 9479.246181261968,
-      "required_pages_per_second_upper_estimate": 0.003781413029065728
+      "required_token_positions_per_second": 9490.538295973818,
+      "required_pages_per_second_upper_estimate": 0.0037859176224564455
     },
     {
       "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": 18958.492362523935,
-      "required_pages_per_second_upper_estimate": 0.00438571355000017
+      "required_token_positions_per_second": 18981.076591947636,
+      "required_pages_per_second_upper_estimate": 0.004390938014008481
     },
     {
       "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": 28437.738543785905,
-      "required_pages_per_second_upper_estimate": 0.0017317090760333052
+      "required_token_positions_per_second": 28471.614887921456,
+      "required_pages_per_second_upper_estimate": 0.0017337719676557186
     },
     {
       "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": 28437.738543785905,
-      "required_pages_per_second_upper_estimate": 0.002189215833799277
+      "required_token_positions_per_second": 28471.614887921456,
+      "required_pages_per_second_upper_estimate": 0.002191823728546554
     },
     {
       "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": 28437.738543785905,
-      "required_pages_per_second_upper_estimate": 0.3531015377253425
+      "required_token_positions_per_second": 28471.614887921456,
+      "required_pages_per_second_upper_estimate": 0.3535221685426755
     }
   ],
   "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": 0.0396436104766309,
-    "producer_p95_seconds": 0.14225521000116714,
-    "gpu_update_seconds": 0.9722292072356621,
-    "producer_batches_per_second": 25.224745879023295,
-    "gpu_batches_per_second": 1.0285640387654045,
-    "producer_headroom": 24.524234688683855,
+    "producer_batch_seconds": 0.03883915171878982,
+    "producer_p95_seconds": 0.13499396800034447,
+    "gpu_update_seconds": 0.9710724210353499,
+    "producer_batches_per_second": 25.747215264647874,
+    "gpu_batches_per_second": 1.0297893116291037,
+    "producer_headroom": 25.002410661959928,
     "sustainable": true,
-    "buffer_seconds": 60.278210848611046,
+    "buffer_seconds": 60.206490104191694,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 67,
+    "upload_pages_per_commit": 68,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790485060735632541-7579de36/train.jsonl

@@ -12,3 +12,4 @@
 {"step": 86200, "loss": 0.6316159069538116, "lr": 0.0003, "grad_norm": 0.1343899664547231, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12614745751, "tokens_per_second": 149288.384666239, "data_wait_s": 9.057999704964459e-06, "gpu_peak_gb": 85.46991729736328}
 {"step": 86300, "loss": 0.6972736828029156, "lr": 0.0003, "grad_norm": 0.1529532199021092, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12629443000, "tokens_per_second": 150473.8301626082, "data_wait_s": 1.6077001419034787e-05, "gpu_peak_gb": 85.46991729736328}
 {"step": 86400, "loss": 0.6343526057898998, "lr": 0.0003, "grad_norm": 0.14067437059617866, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12644099192, "tokens_per_second": 150297.99245291398, "data_wait_s": 8.428000001003966e-06, "gpu_peak_gb": 85.46991729736328}
+{"step": 86500, "loss": 0.6315119042992592, "lr": 0.0003, "grad_norm": 0.13389635879207812, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12658762758, "tokens_per_second": 150075.29958827473, "data_wait_s": 8.196999260690063e-06, "gpu_peak_gb": 85.46991729736328}

runs/void-byte-v1/events.out.tfevents.1790485060.0b09bbb2-f535-42d2-bd00-cbae7fb94f51-r2qrq.28895.0.1790485060735632541-7579de36

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:084dbfcacb8e41a4e1f3442fa30e0992a0315989bfa6926908fb637258e061d8
-size 29817
+oid sha256:683d287a7b06bee1fb92e7ad615cf5a405b875b9b61164932517e1f91d6fd7f1
+size 31935