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": 76101.77691894672,
-      "required_pages_per_second_upper_estimate": 0.006596309145144079
+      "required_token_positions_per_second": 76258.76696444966,
+      "required_pages_per_second_upper_estimate": 0.006609916644400641
     },
     {
       "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": 9512.72211486834,
-      "required_pages_per_second_upper_estimate": 0.0037947670794911204
+      "required_token_positions_per_second": 9532.345870556208,
+      "required_pages_per_second_upper_estimate": 0.0038025952890363046
     },
     {
       "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": 19025.44422973668,
-      "required_pages_per_second_upper_estimate": 0.00440120168616722
+      "required_token_positions_per_second": 19064.691741112416,
+      "required_pages_per_second_upper_estimate": 0.004410280907191388
     },
     {
       "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": 28538.16634460502,
-      "required_pages_per_second_upper_estimate": 0.0017378246021992395
+      "required_token_positions_per_second": 28597.037611668624,
+      "required_pages_per_second_upper_estimate": 0.0017414095534897458
     },
     {
       "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": 28538.16634460502,
-      "required_pages_per_second_upper_estimate": 0.002196947043908278
+      "required_token_positions_per_second": 28597.037611668624,
+      "required_pages_per_second_upper_estimate": 0.002201479117012932
     },
     {
       "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": 28538.16634460502,
-      "required_pages_per_second_upper_estimate": 0.3543485149012878
+      "required_token_positions_per_second": 28597.037611668624,
+      "required_pages_per_second_upper_estimate": 0.3550794990087615
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,20 +155,20 @@
   "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
   "estimated_fresh_generation_batch_seconds": 0.1434574083060457,
   "selected": {
-    "prefetch_batches": 62,
+    "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.04963178421875103,
-    "producer_p95_seconds": 0.12450049300002775,
-    "gpu_update_seconds": 0.9688078647430934,
-    "producer_batches_per_second": 20.14837902245306,
-    "gpu_batches_per_second": 1.032196410033457,
-    "producer_headroom": 19.519908058777283,
+    "producer_batch_seconds": 0.054777027656245814,
+    "producer_p95_seconds": 0.1287122990000853,
+    "gpu_update_seconds": 0.9668134292594914,
+    "producer_batches_per_second": 18.255828086830803,
+    "gpu_batches_per_second": 1.0343257238016719,
+    "producer_headroom": 17.64997975660063,
     "sustainable": true,
-    "buffer_seconds": 60.06608761407179,
+    "buffer_seconds": 60.90924604334796,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 50,
+    "upload_pages_per_commit": 48,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790434090210070261-a05fc3fb/train.jsonl

@@ -1,3 +1,4 @@
 {"step": 40100, "loss": 0.6127832569181919, "lr": 0.0003, "grad_norm": 0.14283070884692345, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 5868007171, "tokens_per_second": 46680.67316178548, "data_wait_s": 8.534000016879872e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 40200, "loss": 0.6159361731261015, "lr": 0.0003, "grad_norm": 0.12924330763801173, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 5882491430, "tokens_per_second": 149490.06081221905, "data_wait_s": 9.717000011733035e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 40300, "loss": 0.6501443609595299, "lr": 0.0003, "grad_norm": 0.14097667619387938, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 5897110808, "tokens_per_second": 150874.0284524783, "data_wait_s": 9.098999953494058e-06, "gpu_peak_gb": 85.46427917480469}
+{"step": 40400, "loss": 0.6147063747048378, "lr": 0.0003, "grad_norm": 0.12158116090798718, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 5911786558, "tokens_per_second": 151573.0581806831, "data_wait_s": 8.046999937505461e-06, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790434090.372bd512-57cc-4928-b44a-3b375a0fa071-5spg5.451.0.1790434090210070261-a05fc3fb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c85ad2590f40bb7e0703c07471311bbcc93d928438780b178400228963374a5d
-size 6519
+oid sha256:93edea663a35169a2a4a4f29c6ef7e6b4d38dd04fe9ba3838fef747c9a0ae51e
+size 8637