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": 76878.09710428426,
-      "required_pages_per_second_upper_estimate": 0.0053186051394358775
+      "required_token_positions_per_second": 76874.95943805526,
+      "required_pages_per_second_upper_estimate": 0.005318388068405767
     },
     {
       "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": 9609.762138035532,
-      "required_pages_per_second_upper_estimate": 0.00402340673254256
+      "required_token_positions_per_second": 9609.369929756907,
+      "required_pages_per_second_upper_estimate": 0.004023242523126539
     },
     {
       "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": 19219.524276071064,
-      "required_pages_per_second_upper_estimate": 0.0040029200537660585
+      "required_token_positions_per_second": 19218.739859513815,
+      "required_pages_per_second_upper_estimate": 0.00400275668048364
     },
     {
       "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": 28829.2864141066,
-      "required_pages_per_second_upper_estimate": 0.0017282717007541637
+      "required_token_positions_per_second": 28828.109789270722,
+      "required_pages_per_second_upper_estimate": 0.001728201163891855
     },
     {
       "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": 28829.2864141066,
-      "required_pages_per_second_upper_estimate": 0.0022413502120671902
+      "required_token_positions_per_second": 28828.109789270722,
+      "required_pages_per_second_upper_estimate": 0.0022412587346616266
     },
     {
       "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": 28829.2864141066,
-      "required_pages_per_second_upper_estimate": 0.36979113164411176
+      "required_token_positions_per_second": 28828.109789270722,
+      "required_pages_per_second_upper_estimate": 0.3697760391640785
     }
   ],
   "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": 720750,
-    "producer_batch_seconds": 0.05557546420310189,
-    "producer_p95_seconds": 0.15498676200013506,
-    "gpu_update_seconds": 0.9590247778894528,
-    "producer_batches_per_second": 17.9935519089049,
-    "gpu_batches_per_second": 1.0427259264361473,
-    "producer_headroom": 17.256262122879864,
+    "producer_batch_seconds": 0.05721420563280333,
+    "producer_p95_seconds": 0.1434586870000203,
+    "gpu_update_seconds": 0.9590639206698894,
+    "producer_batches_per_second": 17.47817677340359,
+    "gpu_batches_per_second": 1.0426833691142479,
+    "producer_headroom": 16.762688742461844,
     "sustainable": true,
-    "buffer_seconds": 60.41856100703553,
+    "buffer_seconds": 60.42102700220303,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791209746770022608-f3fe5fb7/train.jsonl

@@ -37,3 +37,4 @@
 {"step": 723700, "loss": 0.5550183653831482, "lr": 5e-05, "grad_norm": 0.22080920255256617, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105929663900, "tokens_per_second": 150692.5883327505, "data_wait_s": 9.963000138668576e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 723800, "loss": 0.5272859334945679, "lr": 5e-05, "grad_norm": 0.21699936786714963, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105944292649, "tokens_per_second": 151765.92871209202, "data_wait_s": 1.1854999684146605e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 723900, "loss": 0.5037689805030823, "lr": 5e-05, "grad_norm": 0.2189082987368743, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105958933471, "tokens_per_second": 152338.83299431106, "data_wait_s": 1.313800021307543e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 724000, "loss": 0.5868327021598816, "lr": 5e-05, "grad_norm": 0.21152553427254808, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105973600123, "tokens_per_second": 152394.86108512752, "data_wait_s": 1.3791000128549058e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791209746.ebd20fd8-0561-42d8-bf2b-56d08e7106ef-6qqk2.329.0.1791209746770022608-f3fe5fb7

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2625272b76d28d74a791533cdb019702ef6113cef8fe60774876a7fe70a78360
-size 82767
+oid sha256:d1215670fe023d98f95633236f9812373cdd2e6447052fb8d7528b4ea82d2a65
+size 84885