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": 76852.6698112814,
-      "required_pages_per_second_upper_estimate": 0.005316846020306492
+      "required_token_positions_per_second": 76683.03735690395,
+      "required_pages_per_second_upper_estimate": 0.0053051104535111375
     },
     {
       "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": 9606.583726410176,
-      "required_pages_per_second_upper_estimate": 0.004022075997967805
+      "required_token_positions_per_second": 9585.379669612994,
+      "required_pages_per_second_upper_estimate": 0.0040131983021778825
     },
     {
       "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": 19213.16745282035,
-      "required_pages_per_second_upper_estimate": 0.004001596095123638
+      "required_token_positions_per_second": 19170.759339225988,
+      "required_pages_per_second_upper_estimate": 0.003992763603439095
     },
     {
       "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": 28819.751179230527,
-      "required_pages_per_second_upper_estimate": 0.0017277000779828038
+      "required_token_positions_per_second": 28756.139008838985,
+      "required_pages_per_second_upper_estimate": 0.001723886625497991
     },
     {
       "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": 28819.751179230527,
-      "required_pages_per_second_upper_estimate": 0.002240608889496641
+      "required_token_positions_per_second": 28756.139008838985,
+      "required_pages_per_second_upper_estimate": 0.0022356633230490664
     },
     {
       "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": 28819.751179230527,
-      "required_pages_per_second_upper_estimate": 0.3696688238892591
+      "required_token_positions_per_second": 28756.139008838985,
+      "required_pages_per_second_upper_estimate": 0.3688528752689035
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,24 +158,24 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.103063767156101,
-    "producer_p95_seconds": 0.3022905619982339,
-    "gpu_update_seconds": 0.9593420785646835,
-    "producer_batches_per_second": 9.702730916922473,
-    "gpu_batches_per_second": 1.042381046702493,
-    "producer_headroom": 9.308238045594221,
+    "producer_batch_seconds": 0.1041239505859437,
+    "producer_p95_seconds": 0.307868795000104,
+    "gpu_update_seconds": 0.9614642630396811,
+    "producer_batches_per_second": 9.603938329007235,
+    "gpu_batches_per_second": 1.0400802592896043,
+    "producer_headroom": 9.233843487777488,
     "sustainable": true,
-    "buffer_seconds": 60.43855094957506,
+    "buffer_seconds": 60.57224857149991,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 81,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 284002.116762901,
+    "generation_bytes_per_second": 295078.78964297683,
     "upload_bytes_per_second": 5436700.3416867545,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 60.940243026402214,
+    "estimated_full_shards_per_hour": 63.31703917471865,
     "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
   },
   "upload_measurements": [

metrics/void-byte-v1/1790868774355361490-de221e1d/train.jsonl

@@ -306,3 +306,4 @@
 {"step": 470800, "loss": 0.5753962397575378, "lr": 0.0003, "grad_norm": 0.18389009604001882, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 68910712220, "tokens_per_second": 152375.92854890318, "data_wait_s": 1.5951998648233712e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 470900, "loss": 0.6000497341156006, "lr": 0.0003, "grad_norm": 0.1599101394922151, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 68925446745, "tokens_per_second": 147862.43837226127, "data_wait_s": 1.0037998436018825e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 471000, "loss": 0.5956321358680725, "lr": 0.0003, "grad_norm": 0.16470233697343703, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 68940085663, "tokens_per_second": 152234.74635804413, "data_wait_s": 1.0285999451298267e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 471100, "loss": 0.55875563621521, "lr": 0.0003, "grad_norm": 0.14749087728292748, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 68954730461, "tokens_per_second": 151707.98058538302, "data_wait_s": 1.2439999409252778e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1790868774.f92f6aaf-743a-4c24-a54a-850bd75954cb-8js8b.612.0.1790868774355361490-de221e1d

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e5fce9edbd4e16152cdad01e39ce5e764ce6faaf0ab7313eb3f0272e3acdd97a
-size 657207
+oid sha256:a0d665000f3ffb50a9b4e7713e0488a252235a7a1fbdad3b0e8bd4dad47278c9
+size 659325