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": 76666.80380649082,
-      "required_pages_per_second_upper_estimate": 0.0053039873788265305
+      "required_token_positions_per_second": 76570.90204746342,
+      "required_pages_per_second_upper_estimate": 0.005297352672614277
     },
     {
       "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": 9583.350475811352,
-      "required_pages_per_second_upper_estimate": 0.00401234872110752
+      "required_token_positions_per_second": 9571.362755932927,
+      "required_pages_per_second_upper_estimate": 0.004007329713126481
     },
     {
       "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": 19166.700951622704,
-      "required_pages_per_second_upper_estimate": 0.00399191834832821
+      "required_token_positions_per_second": 19142.725511865854,
+      "required_pages_per_second_upper_estimate": 0.003986924896501722
     },
     {
       "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": 28750.051427434057,
-      "required_pages_per_second_upper_estimate": 0.0017235216842879642
+      "required_token_positions_per_second": 28714.08826779878,
+      "required_pages_per_second_upper_estimate": 0.0017213657477803946
     },
     {
       "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": 28750.051427434057,
-      "required_pages_per_second_upper_estimate": 0.0022351900403712738
+      "required_token_positions_per_second": 28714.08826779878,
+      "required_pages_per_second_upper_estimate": 0.002232394062894853
     },
     {
       "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": 28750.051427434057,
-      "required_pages_per_second_upper_estimate": 0.3687747903109767
+      "required_token_positions_per_second": 28714.08826779878,
+      "required_pages_per_second_upper_estimate": 0.3683134935134077
     }
   ],
   "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": 0.04395623965625095,
-    "producer_p95_seconds": 0.11782706700000745,
-    "gpu_update_seconds": 0.9616678450048806,
-    "producer_batches_per_second": 22.749898713362565,
-    "gpu_batches_per_second": 1.0398600776705027,
-    "producer_headroom": 21.877846069758682,
+    "producer_batch_seconds": 0.04866712878906565,
+    "producer_p95_seconds": 0.11993603400003394,
+    "gpu_update_seconds": 0.962872292588362,
+    "producer_batches_per_second": 20.547750090913034,
+    "gpu_batches_per_second": 1.0385593268156388,
+    "producer_headroom": 19.784859237570156,
     "sustainable": true,
-    "buffer_seconds": 60.585074235307474,
+    "buffer_seconds": 60.660954433066806,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 61,
+    "upload_pages_per_commit": 64,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -2,3 +2,4 @@
 {"step": 720200, "loss": 0.5311113595962524, "lr": 5e-05, "grad_norm": 0.2152443446279208, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105416991235, "tokens_per_second": 150859.00996360285, "data_wait_s": 8.567000008952164e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 720300, "loss": 0.5242825150489807, "lr": 5e-05, "grad_norm": 0.20403188480716858, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105431760929, "tokens_per_second": 153686.81241369364, "data_wait_s": 9.222000016961829e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 720400, "loss": 0.5264872312545776, "lr": 5e-05, "grad_norm": 0.20526975189012117, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105446489981, "tokens_per_second": 153182.13289004244, "data_wait_s": 1.7705999994177546e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 720500, "loss": 0.5409865379333496, "lr": 5e-05, "grad_norm": 0.22716974496486186, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105461179941, "tokens_per_second": 152853.1822943083, "data_wait_s": 1.070100006472785e-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 sha256:42cd5ca4ce1b92612e542e223c344cd4dea26d8ec93a456e6b52901595399acf
-size 8637
+oid sha256:f723fa0b2eb61d66c8eed58d0b4464c94dada34b4f0cc8a9c747fbf6acf5b0ab
+size 10755