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": 77352.94445568118,
-      "required_pages_per_second_upper_estimate": 0.00535145618100317
+      "required_token_positions_per_second": 77214.57309422207,
+      "required_pages_per_second_upper_estimate": 0.005341883329151636
     },
     {
       "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": 9669.118056960147,
-      "required_pages_per_second_upper_estimate": 0.004048257816303761
+      "required_token_positions_per_second": 9651.821636777759,
+      "required_pages_per_second_upper_estimate": 0.004041016166363721
     },
     {
       "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": 19338.236113920295,
-      "required_pages_per_second_upper_estimate": 0.004027644598948363
+      "required_token_positions_per_second": 19303.643273555517,
+      "required_pages_per_second_upper_estimate": 0.004020439822574928
     },
     {
       "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": 29007.35417088044,
-      "required_pages_per_second_upper_estimate": 0.0017389465908790342
+      "required_token_positions_per_second": 28955.464910333274,
+      "required_pages_per_second_upper_estimate": 0.0017358359089395447
     },
     {
       "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": 29007.35417088044,
-      "required_pages_per_second_upper_estimate": 0.0022551941969190695
+      "required_token_positions_per_second": 28955.464910333274,
+      "required_pages_per_second_upper_estimate": 0.0022511600351482635
     },
     {
       "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": 29007.35417088044,
-      "required_pages_per_second_upper_estimate": 0.3720751936337456
+      "required_token_positions_per_second": 28955.464910333274,
+      "required_pages_per_second_upper_estimate": 0.37140961391379373
     }
   ],
   "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": 0.04784228773436894,
-    "producer_p95_seconds": 0.12312903399993047,
-    "gpu_update_seconds": 0.9531376021793448,
-    "producer_batches_per_second": 20.902010488131822,
-    "gpu_batches_per_second": 1.0491664558333493,
-    "producer_headroom": 19.922492157385484,
+    "producer_batch_seconds": 0.04850933461717588,
+    "producer_p95_seconds": 0.11361886500003493,
+    "gpu_update_seconds": 0.9548456598993621,
+    "producer_batches_per_second": 20.61458908665233,
+    "gpu_batches_per_second": 1.047289674129531,
+    "producer_headroom": 19.683750919998733,
     "sustainable": true,
-    "buffer_seconds": 60.047668937298724,
+    "buffer_seconds": 60.15527657365981,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 58,
+    "upload_pages_per_commit": 62,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791463982175382982-d4a3b372/train.jsonl

@@ -6,3 +6,4 @@
 {"step": 925600, "loss": 0.5387893319129944, "lr": 3e-05, "grad_norm": 0.337824049911127, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135481383545, "tokens_per_second": 152456.86317038405, "data_wait_s": 7.531999926868593e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 925700, "loss": 0.5608165860176086, "lr": 3e-05, "grad_norm": 0.3066571247254951, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135496029460, "tokens_per_second": 153616.88852898325, "data_wait_s": 1.317600003858388e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 925800, "loss": 0.5568015575408936, "lr": 3e-05, "grad_norm": 0.3269186436657379, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135510638316, "tokens_per_second": 153281.01109958353, "data_wait_s": 1.1348000043653883e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 925900, "loss": 0.5448195934295654, "lr": 3e-05, "grad_norm": 0.35925442046433215, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135525342509, "tokens_per_second": 153838.81587312397, "data_wait_s": 6.365999979607295e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791463982.2e076f0c-3fe2-4316-8f17-4e30b2e706e2-b5972.360.0.1791463982175382982-d4a3b372

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dea4b4ce4e974751325abccabf592b7462f7eae769b919af2e2c048904b6133d
-size 17109
+oid sha256:e26d0b7fac8d9d48b7a0f25dd3f3c3309760138136f180ce59c4983dddbf1146
+size 19227