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": 77007.26741681226,
-      "required_pages_per_second_upper_estimate": 0.005327541441372982
+      "required_token_positions_per_second": 77148.26262232814,
+      "required_pages_per_second_upper_estimate": 0.005337295816844521
     },
     {
       "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": 9625.908427101533,
-      "required_pages_per_second_upper_estimate": 0.004030166846601637
+      "required_token_positions_per_second": 9643.532827791018,
+      "required_pages_per_second_upper_estimate": 0.004037545815130987
     },
     {
       "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": 19251.816854203065,
-      "required_pages_per_second_upper_estimate": 0.0040096457461784
+      "required_token_positions_per_second": 19287.065655582035,
+      "required_pages_per_second_upper_estimate": 0.0040169871419322364
     },
     {
       "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": 28877.725281304596,
-      "required_pages_per_second_upper_estimate": 0.001731175536880816
+      "required_token_positions_per_second": 28930.598483373054,
+      "required_pages_per_second_upper_estimate": 0.001734345204612115
     },
     {
       "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": 28877.725281304596,
-      "required_pages_per_second_upper_estimate": 0.0022451161209317747
+      "required_token_positions_per_second": 28930.598483373054,
+      "required_pages_per_second_upper_estimate": 0.0022492267798279596
     },
     {
       "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": 28877.725281304596,
-      "required_pages_per_second_upper_estimate": 0.3704124534229242
+      "required_token_positions_per_second": 28930.598483373054,
+      "required_pages_per_second_upper_estimate": 0.3710906540882371
     }
   ],
   "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": 0.055020548023431814,
-    "producer_p95_seconds": 0.1732370820000142,
-    "gpu_update_seconds": 0.9574161306222856,
-    "producer_batches_per_second": 18.17502798362034,
-    "gpu_batches_per_second": 1.0444779109268156,
-    "producer_headroom": 17.401064966029548,
+    "producer_batch_seconds": 0.045813664437496016,
+    "producer_p95_seconds": 0.1197988870003428,
+    "gpu_update_seconds": 0.955666368806363,
+    "producer_batches_per_second": 21.82754888258958,
+    "gpu_batches_per_second": 1.0463902807932963,
+    "producer_headroom": 20.859854380567768,
     "sustainable": true,
-    "buffer_seconds": 60.31721622920399,
+    "buffer_seconds": 60.20698123480087,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 65,
+    "upload_pages_per_commit": 66,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791300814123050567-72be911d/train.jsonl

@@ -10,3 +10,4 @@
 {"step": 801000, "loss": 0.5693171620368958, "lr": 3e-05, "grad_norm": 0.26937283032833953, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117242932024, "tokens_per_second": 151567.21791711316, "data_wait_s": 1.225300002261065e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 801100, "loss": 0.5213466286659241, "lr": 3e-05, "grad_norm": 0.28487558318618295, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117257534817, "tokens_per_second": 152277.8665797183, "data_wait_s": 1.150700018115458e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 801200, "loss": 0.5780178308486938, "lr": 3e-05, "grad_norm": 0.2919761416088437, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117272170109, "tokens_per_second": 152201.3480068655, "data_wait_s": 2.256599964312045e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 801300, "loss": 0.5488638877868652, "lr": 3e-05, "grad_norm": 0.27684681061972655, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117286883998, "tokens_per_second": 153382.37674826293, "data_wait_s": 9.282000064558815e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791300814.0c9d201b-2dbc-40fb-b739-4ee2350a0f54-7b2tx.840.0.1791300814123050567-72be911d

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dd1fe4b7f78cd9707b4600773f6b092e2b51eb5bd98c55b127eb3401af319bce
-size 25581
+oid sha256:6f70c74ee65d378583ebfed433a0c05ca76fdfca152d2f3fbd821e1d079d8062
+size 27699