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": 77642.75469030903,
-      "required_pages_per_second_upper_estimate": 0.005371505925487114
+      "required_token_positions_per_second": 77741.3088934432,
+      "required_pages_per_second_upper_estimate": 0.005378324133937193
     },
     {
       "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": 9705.344336288628,
-      "required_pages_per_second_upper_estimate": 0.004063425002967861
+      "required_token_positions_per_second": 9717.6636116804,
+      "required_pages_per_second_upper_estimate": 0.004068582826318672
     },
     {
       "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": 19410.688672577257,
-      "required_pages_per_second_upper_estimate": 0.004042734556214147
+      "required_token_positions_per_second": 19435.3272233608,
+      "required_pages_per_second_upper_estimate": 0.0040478661165800804
     },
     {
       "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": 29116.033008865885,
-      "required_pages_per_second_upper_estimate": 0.0017454617212732842
+      "required_token_positions_per_second": 29152.990835041204,
+      "required_pages_per_second_upper_estimate": 0.0017476772865211616
     },
     {
       "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": 29116.033008865885,
-      "required_pages_per_second_upper_estimate": 0.002263643498544749
+      "required_token_positions_per_second": 29152.990835041204,
+      "required_pages_per_second_upper_estimate": 0.0022665168069695824
     },
     {
       "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": 29116.033008865885,
-      "required_pages_per_second_upper_estimate": 0.3734692090771781
+      "required_token_positions_per_second": 29152.990835041204,
+      "required_pages_per_second_upper_estimate": 0.37394326438913306
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.03050388856249242,
-    "producer_p95_seconds": 0.08444445799977984,
-    "gpu_update_seconds": 0.9495799098586382,
-    "producer_batches_per_second": 32.782705652472124,
-    "gpu_batches_per_second": 1.0530972587118737,
-    "producer_headroom": 31.129798678396746,
+    "producer_batch_seconds": 0.029110802624987997,
+    "producer_p95_seconds": 0.073699526999917,
+    "gpu_update_seconds": 0.9483761085249531,
+    "producer_batches_per_second": 34.35150905601018,
+    "gpu_batches_per_second": 1.0544339856424045,
+    "producer_headroom": 32.578150480498614,
     "sustainable": true,
-    "buffer_seconds": 60.77311423095284,
+    "buffer_seconds": 60.696070945597,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 50,
+    "upload_pages_per_commit": 61,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790914434605211575-090606d4/train.jsonl

@@ -2,3 +2,4 @@
 {"step": 485200, "loss": 0.5576065182685852, "lr": 0.0003, "grad_norm": 0.14968418877833817, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 71016926460, "tokens_per_second": 153832.48290073316, "data_wait_s": 1.3774999843008118e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 485300, "loss": 0.5296840071678162, "lr": 0.0003, "grad_norm": 0.16333932344688673, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 71031548750, "tokens_per_second": 153642.53780209355, "data_wait_s": 7.868000011512777e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 485400, "loss": 0.5068361163139343, "lr": 0.0003, "grad_norm": 0.16174502470969496, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 71046213893, "tokens_per_second": 154217.95321105205, "data_wait_s": 1.3796000075672055e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 485500, "loss": 0.5999634265899658, "lr": 0.0003, "grad_norm": 0.1477095802620134, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 71060987918, "tokens_per_second": 153889.3690089966, "data_wait_s": 1.7192000086652115e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1790914434.4dec8a9d-3312-413a-9119-9058bec1adb7-bx5st.8214.0.1790914434605211575-090606d4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0bb11f27bd6b161ee05ec20163a7c080d13b092e94c7559c7111c7e8a4b3659c
-size 8637
+oid sha256:c7e39c08856e0a2be74956d90b85487d21d3b88e13bbf350ef0d96d6ce945944
+size 10755