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": 76301.26309665144,
-      "required_pages_per_second_upper_estimate": 0.006613600101434429
+      "required_token_positions_per_second": 76376.41714288165,
+      "required_pages_per_second_upper_estimate": 0.006620114263685474
     },
     {
       "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": 9537.65788708143,
-      "required_pages_per_second_upper_estimate": 0.00380471433184994
+      "required_token_positions_per_second": 9547.052142860206,
+      "required_pages_per_second_upper_estimate": 0.0038084618409367343
     },
     {
       "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": 19075.31577416286,
-      "required_pages_per_second_upper_estimate": 0.004412738590261012
+      "required_token_positions_per_second": 19094.104285720412,
+      "required_pages_per_second_upper_estimate": 0.004417084981743342
     },
     {
       "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": 28612.973661244287,
-      "required_pages_per_second_upper_estimate": 0.0017423799752989133
+      "required_token_positions_per_second": 28641.156428580616,
+      "required_pages_per_second_upper_estimate": 0.001744096157966143
     },
     {
       "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": 28612.973661244287,
-      "required_pages_per_second_upper_estimate": 0.002202705918223075
+      "required_token_positions_per_second": 28641.156428580616,
+      "required_pages_per_second_upper_estimate": 0.002204875505667519
     },
     {
       "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": 28612.973661244287,
-      "required_pages_per_second_upper_estimate": 0.3552773714099642
+      "required_token_positions_per_second": 28641.156428580616,
+      "required_pages_per_second_upper_estimate": 0.35562730705862666
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,14 +158,14 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 690670,
-    "producer_batch_seconds": 0.0569888585546785,
-    "producer_p95_seconds": 0.1384437749998142,
-    "gpu_update_seconds": 0.9662749606989879,
-    "producer_batches_per_second": 17.547289511695354,
-    "gpu_batches_per_second": 1.034902114483662,
-    "producer_headroom": 16.955506483287188,
+    "producer_batch_seconds": 0.05107486564842034,
+    "producer_p95_seconds": 0.14190697799995178,
+    "gpu_update_seconds": 0.9653241505433922,
+    "producer_batches_per_second": 19.579101918419404,
+    "gpu_batches_per_second": 1.0359214564735466,
+    "producer_headroom": 18.90017992780071,
     "sustainable": true,
-    "buffer_seconds": 60.875322524036235,
+    "buffer_seconds": 60.8154214842337,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 83,

metrics/void-byte-v1/1790696019647648547-3fdea0a8/train.jsonl

@@ -22,3 +22,4 @@
 {"step": 277200, "loss": 0.6728517189621925, "lr": 0.0003, "grad_norm": 0.1942734196814653, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40574057787, "tokens_per_second": 149074.59816718844, "data_wait_s": 9.401000170328189e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 277300, "loss": 0.5808306690305471, "lr": 0.0003, "grad_norm": 0.15871520061710287, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40588611599, "tokens_per_second": 149994.4769097146, "data_wait_s": 1.2185999821667792e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 277400, "loss": 0.567566778510809, "lr": 0.0003, "grad_norm": 0.14623475431320984, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40603211126, "tokens_per_second": 150490.46349487503, "data_wait_s": 1.1923999863938661e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 277500, "loss": 0.6089343391358852, "lr": 0.0003, "grad_norm": 0.1495398169538208, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40617975581, "tokens_per_second": 152355.8195530897, "data_wait_s": 1.25619999380433e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790696019.3e6d470a-8cc7-4d7e-a224-51a0f9a2ce55-85btc.22407.0.1790696019647648547-3fdea0a8

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ec7094213c591c6642ea0ef59f1fe0b057849f176883426c809bea8f9795e7b4
-size 50997
+oid sha256:375776c4f7ac3344351966587196605d5a706b896ded06211f1fc2bec92c46e0
+size 53115