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": 77519.99477366949,
-      "required_pages_per_second_upper_estimate": 0.005363013109611741
+      "required_token_positions_per_second": 77607.04556905093,
+      "required_pages_per_second_upper_estimate": 0.005369035485621902
     },
     {
       "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": 9689.999346708686,
-      "required_pages_per_second_upper_estimate": 0.00405700037627056
+      "required_token_positions_per_second": 9700.880696131366,
+      "required_pages_per_second_upper_estimate": 0.0040615561700454205
     },
     {
       "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": 19379.998693417372,
-      "required_pages_per_second_upper_estimate": 0.00403634264290432
+      "required_token_positions_per_second": 19401.76139226273,
+      "required_pages_per_second_upper_estimate": 0.004040875239152845
     },
     {
       "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": 29069.998040126062,
-      "required_pages_per_second_upper_estimate": 0.001742701995188659
+      "required_token_positions_per_second": 29102.6420883941,
+      "required_pages_per_second_upper_estimate": 0.0017446589560377523
     },
     {
       "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": 29069.998040126062,
-      "required_pages_per_second_upper_estimate": 0.0022600644822116556
+      "required_token_positions_per_second": 29102.6420883941,
+      "required_pages_per_second_upper_estimate": 0.002262602413378502
     },
     {
       "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": 29069.998040126062,
-      "required_pages_per_second_upper_estimate": 0.3728787219266821
+      "required_token_positions_per_second": 29102.6420883941,
+      "required_pages_per_second_upper_estimate": 0.37329744472741627
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.05154047796091632,
-    "producer_p95_seconds": 0.16849219200003063,
-    "gpu_update_seconds": 0.9510836554524964,
-    "producer_batches_per_second": 19.40222597000964,
-    "gpu_batches_per_second": 1.0514322207800224,
-    "producer_headroom": 18.453139999472125,
+    "producer_batch_seconds": 0.050154552109361816,
+    "producer_p95_seconds": 0.14173550299983617,
+    "gpu_update_seconds": 0.9500168375099456,
+    "producer_batches_per_second": 19.938369658241662,
+    "gpu_batches_per_second": 1.0526129227573098,
+    "producer_headroom": 18.941786887827,
     "sustainable": true,
-    "buffer_seconds": 60.86935394895977,
+    "buffer_seconds": 60.80107760063652,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 85,
+    "upload_pages_per_commit": 88,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791143623306220418-eb05a537/train.jsonl

@@ -35,3 +35,4 @@
 {"step": 688500, "loss": 0.5517596006393433, "lr": 8e-05, "grad_norm": 0.18455588967613337, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100778019023, "tokens_per_second": 153112.0184768985, "data_wait_s": 1.4700000065204222e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 688600, "loss": 0.5771710872650146, "lr": 8e-05, "grad_norm": 0.2004392058658827, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100792624872, "tokens_per_second": 153019.46159830692, "data_wait_s": 1.3870000202587107e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 688700, "loss": 0.5927730202674866, "lr": 8e-05, "grad_norm": 0.2079924289921846, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100807133319, "tokens_per_second": 152234.86601582528, "data_wait_s": 1.4876000022923108e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 688800, "loss": 0.5484572052955627, "lr": 8e-05, "grad_norm": 0.1872351369526101, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100821681614, "tokens_per_second": 152381.7719375346, "data_wait_s": 1.6102000245155068e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791143623.3eb6ddb3-ffce-42e5-bd48-c0d653189e98-8pb7d.369.0.1791143623306220418-eb05a537

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:78b1778ac9e6287d8f38769555da796f0faf3cbf452265203f30054915ab959b
-size 78531
+oid sha256:64ba486ccdb7174a4f7424c660db6d1007b5fc68902727b1cada076fb641a7d2
+size 80649