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": 77214.57309422207,
-      "required_pages_per_second_upper_estimate": 0.005341883329151636
+      "required_token_positions_per_second": 77146.20748894117,
+      "required_pages_per_second_upper_estimate": 0.0053371536379482444
     },
     {
       "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": 9651.821636777759,
-      "required_pages_per_second_upper_estimate": 0.004041016166363721
+      "required_token_positions_per_second": 9643.275936117647,
+      "required_pages_per_second_upper_estimate": 0.004037438259951855
     },
     {
       "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": 19303.643273555517,
-      "required_pages_per_second_upper_estimate": 0.004020439822574928
+      "required_token_positions_per_second": 19286.551872235294,
+      "required_pages_per_second_upper_estimate": 0.004016880134410488
     },
     {
       "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": 28955.464910333274,
-      "required_pages_per_second_upper_estimate": 0.0017358359089395447
+      "required_token_positions_per_second": 28929.82780835294,
+      "required_pages_per_second_upper_estimate": 0.0017342990038214118
     },
     {
       "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": 28955.464910333274,
-      "required_pages_per_second_upper_estimate": 0.0022511600351482635
+      "required_token_positions_per_second": 28929.82780835294,
+      "required_pages_per_second_upper_estimate": 0.002249166863234987
     },
     {
       "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": 28955.464910333274,
-      "required_pages_per_second_upper_estimate": 0.37140961391379373
+      "required_token_positions_per_second": 28929.82780835294,
+      "required_pages_per_second_upper_estimate": 0.3710807686965655
     }
   ],
   "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": 0.04850933461717588,
-    "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,
+    "producer_batch_seconds": 0.054622626296875865,
+    "producer_p95_seconds": 0.1252393499999016,
+    "gpu_update_seconds": 0.9556918272433396,
+    "producer_batches_per_second": 18.307431696252856,
+    "gpu_batches_per_second": 1.0463624062627654,
+    "producer_headroom": 17.496262849924523,
     "sustainable": true,
-    "buffer_seconds": 60.15527657365981,
+    "buffer_seconds": 60.20858511633039,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 62,
+    "upload_pages_per_commit": 64,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -7,3 +7,4 @@
 {"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}
+{"step": 926000, "loss": 0.5181822180747986, "lr": 3e-05, "grad_norm": 0.3358263786675822, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135540055118, "tokens_per_second": 154409.78852788528, "data_wait_s": 1.0637000059432467e-05, "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:e26d0b7fac8d9d48b7a0f25dd3f3c3309760138136f180ce59c4983dddbf1146
-size 19227
+oid sha256:bc3d3715264f81f81f2ff857c7f725ced654724646d8a8ad533f964191c99c44
+size 21345