Update TensorBoard training metrics

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 1062967.6168812835,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 75524.60768576761,
-      "required_pages_per_second_upper_estimate": 0.006546281578834161
+      "required_token_positions_per_second": 75624.81359048597,
+      "required_pages_per_second_upper_estimate": 0.006554967172685605
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 1206672.263377472,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9440.575960720951,
-      "required_pages_per_second_upper_estimate": 0.0037659868999205966
+      "required_token_positions_per_second": 9453.101698810746,
+      "required_pages_per_second_upper_estimate": 0.003770983604121089
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 410042.7581134458,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 18881.151921441902,
-      "required_pages_per_second_upper_estimate": 0.004367822200101116
+      "required_token_positions_per_second": 18906.20339762149,
+      "required_pages_per_second_upper_estimate": 0.004373617418224341
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 2163486.5624068812,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28321.727882162857,
-      "required_pages_per_second_upper_estimate": 0.0017246446354013633
+      "required_token_positions_per_second": 28359.305096432236,
+      "required_pages_per_second_upper_estimate": 0.001726932890597962
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 1546694.2737633225,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28321.727882162857,
-      "required_pages_per_second_upper_estimate": 0.0021802850119294697
+      "required_token_positions_per_second": 28359.305096432236,
+      "required_pages_per_second_upper_estimate": 0.002183177809904313
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 22,
-      "required_token_positions_per_second": 28321.727882162857,
-      "required_pages_per_second_upper_estimate": 0.35166107357069243
+      "required_token_positions_per_second": 28359.305096432236,
+      "required_pages_per_second_upper_estimate": 0.35212765680907204
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,14 +158,14 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.05853812021093674,
-    "producer_p95_seconds": 0.15417635499989046,
-    "gpu_update_seconds": 0.976211625047525,
-    "producer_batches_per_second": 17.082885415462467,
-    "gpu_batches_per_second": 1.024368051293506,
-    "producer_headroom": 16.67651133192928,
+    "producer_batch_seconds": 0.05614438421092771,
+    "producer_p95_seconds": 0.1496411060002174,
+    "gpu_update_seconds": 0.9749181055736897,
+    "producer_batches_per_second": 17.811220375008837,
+    "gpu_batches_per_second": 1.025727180860541,
+    "producer_headroom": 17.364481225959118,
     "sustainable": true,
-    "buffer_seconds": 60.52512075294655,
+    "buffer_seconds": 60.444922545568765,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 78,

metrics/void-byte-v1/1790606918076636751-3952f1c5/train.jsonl

@@ -25,3 +25,4 @@
 {"step": 197500, "loss": 0.5895135682076216, "lr": 0.0003, "grad_norm": 0.13332616485866206, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28906201905, "tokens_per_second": 148764.0805458956, "data_wait_s": 1.8090000139636686e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 197600, "loss": 0.5868651624768972, "lr": 0.0003, "grad_norm": 0.17262299639540224, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28920839583, "tokens_per_second": 149402.8234699391, "data_wait_s": 1.622199988560169e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 197700, "loss": 0.6026234813034534, "lr": 0.0003, "grad_norm": 0.15512142850393992, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28935300702, "tokens_per_second": 147646.94076719318, "data_wait_s": 1.7273000139539363e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 197800, "loss": 0.5235539898276329, "lr": 0.0003, "grad_norm": 0.149242034860401, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28949997669, "tokens_per_second": 150145.07525097489, "data_wait_s": 1.1808000181190437e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790606918.0a7f8bdc-23d3-4ecb-bd0c-4a4bf460026e-rl2cx.1192.0.1790606918076636751-3952f1c5

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:88902ccdc3e33725c26bc397b691369eac28b303fdc192317cdd9af7636e353b
-size 57351
+oid sha256:815a2dd026143da7cfc2b865c5a23487f23a08efa20bd6ee1426d9221c35e08d
+size 59469