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": 76449.06082677413,
-      "required_pages_per_second_upper_estimate": 0.006626410834091542
+      "required_token_positions_per_second": 76309.04021336343,
+      "required_pages_per_second_upper_estimate": 0.006614274202200102
     },
     {
       "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": 9556.132603346767,
-      "required_pages_per_second_upper_estimate": 0.0038120841723898063
+      "required_token_positions_per_second": 9538.63002667043,
+      "required_pages_per_second_upper_estimate": 0.003805102132866774
     },
     {
       "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": 19112.265206693533,
-      "required_pages_per_second_upper_estimate": 0.004421286191189181
+      "required_token_positions_per_second": 19077.26005334086,
+      "required_pages_per_second_upper_estimate": 0.004413188365030698
     },
     {
       "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": 28668.3978100403,
-      "required_pages_per_second_upper_estimate": 0.001745755015172555
+      "required_token_positions_per_second": 28615.89008001129,
+      "required_pages_per_second_upper_estimate": 0.001742557569900557
     },
     {
       "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": 28668.3978100403,
-      "required_pages_per_second_upper_estimate": 0.0022069726226212555
+      "required_token_positions_per_second": 28615.89008001129,
+      "required_pages_per_second_upper_estimate": 0.0022029304322129226
     },
     {
       "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": 28668.3978100403,
-      "required_pages_per_second_upper_estimate": 0.3559655538453171
+      "required_token_positions_per_second": 28615.89008001129,
+      "required_pages_per_second_upper_estimate": 0.35531358357042464
     }
   ],
   "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": 690670,
-    "producer_batch_seconds": 0.04854050072657756,
-    "producer_p95_seconds": 0.1356678419997479,
-    "gpu_update_seconds": 0.9644068769799569,
-    "producer_batches_per_second": 20.601353200554566,
-    "gpu_batches_per_second": 1.0369067494950919,
-    "producer_headroom": 19.86808670170787,
+    "producer_batch_seconds": 0.04529537349217705,
+    "producer_p95_seconds": 0.12967073999971035,
+    "gpu_update_seconds": 0.9661764817622298,
+    "producer_batches_per_second": 22.077309952476927,
+    "gpu_batches_per_second": 1.035007598380038,
+    "producer_headroom": 21.33057765665842,
     "sustainable": true,
-    "buffer_seconds": 60.75763324973728,
+    "buffer_seconds": 60.869118351020475,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 78,
+    "upload_pages_per_commit": 82,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -28,3 +28,4 @@
 {"step": 277800, "loss": 0.6053993627429008, "lr": 0.0003, "grad_norm": 0.13589386939837403, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40662069546, "tokens_per_second": 153458.26789603208, "data_wait_s": 1.2136999885115074e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 277900, "loss": 0.5786179415881634, "lr": 0.0003, "grad_norm": 0.15734794308123032, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40676732679, "tokens_per_second": 151542.69303228127, "data_wait_s": 1.229699955729302e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 278000, "loss": 0.5908274631947279, "lr": 0.0003, "grad_norm": 0.14033135135848648, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40691426002, "tokens_per_second": 151521.56213670966, "data_wait_s": 1.3419999959296547e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 278100, "loss": 0.5920380316674709, "lr": 0.0003, "grad_norm": 0.1494044021047503, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40706113567, "tokens_per_second": 150710.19329023614, "data_wait_s": 1.1837999863928417e-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:f447324edef2c1c0f08a88c091f30553a6a4134de0590c936c1775a5ef95cbf3
-size 63705
+oid sha256:e8ffdaa7178b8b414775cafad880488ced1f924a8370976ffdc5d015f0bbf665
+size 65823