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": 76873.77119012331,
-      "required_pages_per_second_upper_estimate": 0.005318305862656719
+      "required_token_positions_per_second": 76699.19859037672,
+      "required_pages_per_second_upper_estimate": 0.005306228525142012
     },
     {
       "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": 9609.221398765414,
-      "required_pages_per_second_upper_estimate": 0.004023180336302081
+      "required_token_positions_per_second": 9587.39982379709,
+      "required_pages_per_second_upper_estimate": 0.004014044098549147
     },
     {
       "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": 19218.442797530828,
-      "required_pages_per_second_upper_estimate": 0.0040026948103066345
+      "required_token_positions_per_second": 19174.79964759418,
+      "required_pages_per_second_upper_estimate": 0.003993605093122092
     },
     {
       "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": 28827.664196296242,
-      "required_pages_per_second_upper_estimate": 0.001728174451273416
+      "required_token_positions_per_second": 28762.19947139127,
+      "required_pages_per_second_upper_estimate": 0.0017242499409742102
     },
     {
       "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": 28827.664196296242,
-      "required_pages_per_second_upper_estimate": 0.002241224091767817
+      "required_token_positions_per_second": 28762.19947139127,
+      "required_pages_per_second_upper_estimate": 0.002236134497355349
     },
     {
       "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": 28827.664196296242,
-      "required_pages_per_second_upper_estimate": 0.36977032357584233
+      "required_token_positions_per_second": 28762.19947139127,
+      "required_pages_per_second_upper_estimate": 0.3689306123753065
     }
   ],
   "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": 720750,
-    "producer_batch_seconds": 0.05248529882032571,
-    "producer_p95_seconds": 0.14219162699964727,
-    "gpu_update_seconds": 0.9590787450463016,
-    "producer_batches_per_second": 19.052954302943498,
-    "gpu_batches_per_second": 1.0426672524702054,
-    "producer_headroom": 18.273283502291584,
+    "producer_batch_seconds": 0.056021767851554216,
+    "producer_p95_seconds": 0.137696139000127,
+    "gpu_update_seconds": 0.9612616735900352,
+    "producer_batches_per_second": 17.85020427505586,
+    "gpu_batches_per_second": 1.0402994600474273,
+    "producer_headroom": 17.158717235364197,
     "sustainable": true,
-    "buffer_seconds": 60.42196093791701,
+    "buffer_seconds": 60.55948543617222,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 86,

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

@@ -19,3 +19,4 @@
 {"step": 801900, "loss": 0.5472532510757446, "lr": 3e-05, "grad_norm": 0.2634064285754349, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117374697652, "tokens_per_second": 153803.20534128963, "data_wait_s": 9.698000212665647e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 802000, "loss": 0.5650686621665955, "lr": 3e-05, "grad_norm": 0.28666325371970175, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117389459163, "tokens_per_second": 153864.7618857791, "data_wait_s": 1.6775999938545283e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 802100, "loss": 0.48784488439559937, "lr": 3e-05, "grad_norm": 0.2600461222229231, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117404298080, "tokens_per_second": 153931.20470258663, "data_wait_s": 1.785199992809794e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 802200, "loss": 0.5233058333396912, "lr": 3e-05, "grad_norm": 0.26054259389607703, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117418832960, "tokens_per_second": 151372.29255512965, "data_wait_s": 1.4666999959445093e-05, "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:cf02c0e0721dd0c7b32f0968f53040fb4d417369c46403de2ada799fc7c85478
-size 44643
+oid sha256:c842058f448c5a85c0c6db6194f3558d57eaf637d8b285fa80cba4bd10b00454
+size 46761