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": 76016.24589793394,
-      "required_pages_per_second_upper_estimate": 0.006588895533019082
+      "required_token_positions_per_second": 75757.09974805996,
+      "required_pages_per_second_upper_estimate": 0.006566433401547917
     },
     {
       "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": 9502.030737241743,
-      "required_pages_per_second_upper_estimate": 0.00379050212910553
+      "required_token_positions_per_second": 9469.637468507495,
+      "required_pages_per_second_upper_estimate": 0.003777579969884911
     },
     {
       "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": 19004.061474483486,
-      "required_pages_per_second_upper_estimate": 0.004396255162063033
+      "required_token_positions_per_second": 18939.27493701499,
+      "required_pages_per_second_upper_estimate": 0.004381267936823813
     },
     {
       "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": 28506.092211725227,
-      "required_pages_per_second_upper_estimate": 0.00173587145578683
+      "required_token_positions_per_second": 28408.912405522486,
+      "required_pages_per_second_upper_estimate": 0.001729953715452122
     },
     {
       "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": 28506.092211725227,
-      "required_pages_per_second_upper_estimate": 0.002194477888372309
+      "required_token_positions_per_second": 28408.912405522486,
+      "required_pages_per_second_upper_estimate": 0.002186996717879901
     },
     {
       "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": 28506.092211725227,
-      "required_pages_per_second_upper_estimate": 0.35395026151613823
+      "required_token_positions_per_second": 28408.912405522486,
+      "required_pages_per_second_upper_estimate": 0.3527436135629895
     }
   ],
   "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": 0.03939259485161983,
-    "producer_p95_seconds": 0.12816023599953041,
-    "gpu_update_seconds": 0.9698979360148048,
-    "producer_batches_per_second": 25.38548180861662,
-    "gpu_batches_per_second": 1.0310363213152933,
-    "producer_headroom": 24.621326410918634,
+    "producer_batch_seconds": 0.039497209624869356,
+    "producer_p95_seconds": 0.12831051600005594,
+    "gpu_update_seconds": 0.9732157150312248,
+    "producer_batches_per_second": 25.31824423795122,
+    "gpu_batches_per_second": 1.0275214267043724,
+    "producer_headroom": 24.640113169372885,
     "sustainable": true,
-    "buffer_seconds": 60.1336720329179,
+    "buffer_seconds": 60.33937433193594,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 79,

metrics/void-byte-v1/1790485060735632541-7579de36/train.jsonl

@@ -19,3 +19,4 @@
 {"step": 86900, "loss": 0.5960618462413549, "lr": 0.0003, "grad_norm": 0.15141781592707554, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12717146045, "tokens_per_second": 149597.13523622364, "data_wait_s": 7.530999937443994e-06, "gpu_peak_gb": 85.46991729736328}
 {"step": 87000, "loss": 0.6178943142294884, "lr": 0.0003, "grad_norm": 0.13060739307729685, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12731783694, "tokens_per_second": 150387.5653255563, "data_wait_s": 8.342998626176268e-06, "gpu_peak_gb": 85.46991729736328}
 {"step": 87100, "loss": 0.6128315981477499, "lr": 0.0003, "grad_norm": 0.13664678851709444, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12746455790, "tokens_per_second": 150061.36194513095, "data_wait_s": 1.0650999684003182e-05, "gpu_peak_gb": 85.46991729736328}
+{"step": 87200, "loss": 0.583887867629528, "lr": 0.0003, "grad_norm": 0.13447308791749443, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12761166726, "tokens_per_second": 150569.09252748732, "data_wait_s": 8.68499955686275e-06, "gpu_peak_gb": 85.46991729736328}

runs/void-byte-v1/events.out.tfevents.1790485060.0b09bbb2-f535-42d2-bd00-cbae7fb94f51-r2qrq.28895.0.1790485060735632541-7579de36

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f99ffce04d56f44d5853b9ba3e13661862e55bda9357f0ba4d615407aac0cbf
-size 44643
+oid sha256:8b439984bc49b43b50711c3e38ecce0b938c6ab97a72a62f8c10e01e355275d1
+size 46761