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": 77509.06560790465,
-      "required_pages_per_second_upper_estimate": 0.005362257004564971
+      "required_token_positions_per_second": 77740.2551807217,
+      "required_pages_per_second_upper_estimate": 0.00537825123564617
     },
     {
       "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": 9688.63320098808,
-      "required_pages_per_second_upper_estimate": 0.004056428399585709
+      "required_token_positions_per_second": 9717.531897590212,
+      "required_pages_per_second_upper_estimate": 0.004068527680379613
     },
     {
       "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": 19377.26640197616,
-      "required_pages_per_second_upper_estimate": 0.0040357735786524865
+      "required_token_positions_per_second": 19435.063795180424,
+      "required_pages_per_second_upper_estimate": 0.004047811251437176
     },
     {
       "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": 29065.899602964244,
-      "required_pages_per_second_upper_estimate": 0.0017424563001387594
+      "required_token_positions_per_second": 29152.595692770636,
+      "required_pages_per_second_upper_estimate": 0.0017476535983453883
     },
     {
       "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": 29065.899602964244,
-      "required_pages_per_second_upper_estimate": 0.002259745846749444
+      "required_token_positions_per_second": 29152.595692770636,
+      "required_pages_per_second_upper_estimate": 0.002266486086396093
     },
     {
       "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": 29065.899602964244,
-      "required_pages_per_second_upper_estimate": 0.3728261515753293
+      "required_token_positions_per_second": 29152.595692770636,
+      "required_pages_per_second_upper_estimate": 0.3739381959283569
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,14 +158,14 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.04932320940622503,
-    "producer_p95_seconds": 0.13084836600000926,
-    "gpu_update_seconds": 0.9512177630029507,
-    "producer_batches_per_second": 20.274430882305705,
-    "gpu_batches_per_second": 1.0512839844822137,
-    "producer_headroom": 19.28539879002477,
+    "producer_batch_seconds": 0.04865237901560704,
+    "producer_p95_seconds": 0.13437257299983685,
+    "gpu_update_seconds": 0.9483889630746071,
+    "producer_batches_per_second": 20.553979481233863,
+    "gpu_batches_per_second": 1.0544196937489378,
+    "producer_headroom": 19.493167287264132,
     "sustainable": true,
-    "buffer_seconds": 60.877936832188844,
+    "buffer_seconds": 60.69689363677485,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 79,

metrics/void-byte-v1/1791083219122319327-61ae60ac/train.jsonl

@@ -20,3 +20,4 @@
 {"step": 647000, "loss": 0.578724205493927, "lr": 0.0001, "grad_norm": 0.17413717870327522, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94704061158, "tokens_per_second": 151499.552097465, "data_wait_s": 1.2176999916846398e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 647100, "loss": 0.541343629360199, "lr": 0.0001, "grad_norm": 0.18917820900379853, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94718643414, "tokens_per_second": 153102.9512978919, "data_wait_s": 7.906999599072151e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 647200, "loss": 0.5620786547660828, "lr": 0.0001, "grad_norm": 0.19096447251164675, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94733241049, "tokens_per_second": 153358.16760037883, "data_wait_s": 1.4667999948869692e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 647300, "loss": 0.5541459321975708, "lr": 0.0001, "grad_norm": 0.19038575680539732, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94747831707, "tokens_per_second": 152961.1084661129, "data_wait_s": 1.2200999663036782e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791083219.1a73f132-ede3-4b51-9e4f-0cafff989449-8p4r7.435.0.1791083219122319327-61ae60ac

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7e8fb60400b20a3a0eac4da4d48ae8f661af84db89e918d48bc5dc6566abd69f
-size 46761
+oid sha256:78e4428ade4ff78a9a821d5bc45b6677600f856af8bba7bc5b35b4b7664ebeed
+size 48879