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": 76998.1675200524,
-      "required_pages_per_second_upper_estimate": 0.005326911889400467
+      "required_token_positions_per_second": 76566.27054229827,
+      "required_pages_per_second_upper_estimate": 0.005297032254340379
     },
     {
       "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": 9624.77094000655,
-      "required_pages_per_second_upper_estimate": 0.0040296906045084
+      "required_token_positions_per_second": 9570.783817787284,
+      "required_pages_per_second_upper_estimate": 0.004007087323814503
     },
     {
       "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": 19249.5418800131,
-      "required_pages_per_second_upper_estimate": 0.004009171929049735
+      "required_token_positions_per_second": 19141.567635574567,
+      "required_pages_per_second_upper_estimate": 0.003986683741405498
     },
     {
       "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": 28874.312820019648,
-      "required_pages_per_second_upper_estimate": 0.00173097096503731
+      "required_token_positions_per_second": 28712.351453361855,
+      "required_pages_per_second_upper_estimate": 0.001721261628406863
     },
     {
       "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": 28874.312820019648,
-      "required_pages_per_second_upper_estimate": 0.0022448508170767006
+      "required_token_positions_per_second": 28712.351453361855,
+      "required_pages_per_second_upper_estimate": 0.002232259033211821
     },
     {
       "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": 28874.312820019648,
-      "required_pages_per_second_upper_estimate": 0.37036868203357637
+      "required_token_positions_per_second": 28712.351453361855,
+      "required_pages_per_second_upper_estimate": 0.3682912155226569
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.053388736296899,
-    "producer_p95_seconds": 0.12870995699995547,
-    "gpu_update_seconds": 0.9575292812104813,
-    "producer_batches_per_second": 18.7305426080685,
-    "gpu_batches_per_second": 1.044354485677794,
-    "producer_headroom": 17.935043000186127,
+    "producer_batch_seconds": 0.05365086963282906,
+    "producer_p95_seconds": 0.13487341700010802,
+    "gpu_update_seconds": 0.9629305368774583,
+    "producer_batches_per_second": 18.639026857229137,
+    "gpu_batches_per_second": 1.0384965080064328,
+    "producer_headroom": 17.948088138505017,
     "sustainable": true,
-    "buffer_seconds": 60.32434471626032,
+    "buffer_seconds": 60.664623823279875,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 69,
+    "upload_pages_per_commit": 67,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791300814123050567-72be911d/train.jsonl

@@ -15,3 +15,4 @@
 {"step": 801500, "loss": 0.5405973792076111, "lr": 3e-05, "grad_norm": 0.2594121739967866, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117316172213, "tokens_per_second": 151355.2211082398, "data_wait_s": 1.9609999981184956e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 801600, "loss": 0.5029222965240479, "lr": 3e-05, "grad_norm": 0.2854934895432453, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117330622104, "tokens_per_second": 150609.80640364293, "data_wait_s": 1.6166000023076776e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 801700, "loss": 0.5376223921775818, "lr": 3e-05, "grad_norm": 0.2600856098675973, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117345270836, "tokens_per_second": 152675.0416824622, "data_wait_s": 1.4793999980611261e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 801800, "loss": 0.5059638023376465, "lr": 3e-05, "grad_norm": 0.27347432217368667, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117359953415, "tokens_per_second": 153247.0961719905, "data_wait_s": 1.9089000034000492e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791300814.0c9d201b-2dbc-40fb-b739-4ee2350a0f54-7b2tx.840.0.1791300814123050567-72be911d

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bd6746619d0234c8a70f2b6b216bd9d7a694e92033f0f9d483ce33e9b4ce93e1
-size 36171
+oid sha256:03a61fa5f1bffae119bb97b500bd77c0842de47c65b6b0f4e50320aa1953750f
+size 38289