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": 77442.94822254854,
-      "required_pages_per_second_upper_estimate": 0.005357682850432565
+      "required_token_positions_per_second": 77462.78380068325,
+      "required_pages_per_second_upper_estimate": 0.005359055121752806
     },
     {
       "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": 9680.368527818568,
-      "required_pages_per_second_upper_estimate": 0.004052968153515635
+      "required_token_positions_per_second": 9682.847975085406,
+      "required_pages_per_second_upper_estimate": 0.004054006246309513
     },
     {
       "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": 19360.737055637135,
-      "required_pages_per_second_upper_estimate": 0.004032330951718244
+      "required_token_positions_per_second": 19365.69595017081,
+      "required_pages_per_second_upper_estimate": 0.004033363758674765
     },
     {
       "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": 29041.105583455705,
-      "required_pages_per_second_upper_estimate": 0.0017409699365274981
+      "required_token_positions_per_second": 29048.543925256217,
+      "required_pages_per_second_upper_estimate": 0.0017414158537607486
     },
     {
       "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": 29041.105583455705,
-      "required_pages_per_second_upper_estimate": 0.0022578182207900217
+      "required_token_positions_per_second": 29048.543925256217,
+      "required_pages_per_second_upper_estimate": 0.0022583965191472037
     },
     {
       "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": 29041.105583455705,
-      "required_pages_per_second_upper_estimate": 0.37250812051481774
+      "required_token_positions_per_second": 29048.543925256217,
+      "required_pages_per_second_upper_estimate": 0.3726035315767655
     }
   ],
   "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": 0.059583505460967245,
-    "producer_p95_seconds": 0.15197271100032594,
-    "gpu_update_seconds": 0.9520298709202952,
-    "producer_batches_per_second": 16.783168299070507,
-    "gpu_batches_per_second": 1.0503872100497578,
-    "producer_headroom": 15.978077549397687,
+    "producer_batch_seconds": 0.05579623885153495,
+    "producer_p95_seconds": 0.15368040900011692,
+    "gpu_update_seconds": 0.9517860885261613,
+    "producer_batches_per_second": 17.922354993512077,
+    "gpu_batches_per_second": 1.0506562472965935,
+    "producer_headroom": 17.058248156452176,
     "sustainable": true,
-    "buffer_seconds": 60.929911738898895,
+    "buffer_seconds": 60.914309665674324,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 78,
+    "upload_pages_per_commit": 75,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791254470898546162-7724d9f6/train.jsonl

@@ -32,3 +32,4 @@
 {"step": 763200, "loss": 0.5169625878334045, "lr": 3e-05, "grad_norm": 0.22978181003732107, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111709830959, "tokens_per_second": 153816.96861438148, "data_wait_s": 1.4072999874770176e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 763300, "loss": 0.5884107351303101, "lr": 3e-05, "grad_norm": 0.26700905192398544, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111724509274, "tokens_per_second": 153966.58676698315, "data_wait_s": 1.3979000414110487e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 763400, "loss": 0.5867244601249695, "lr": 3e-05, "grad_norm": 0.2573346227700585, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111739120005, "tokens_per_second": 153199.20738341025, "data_wait_s": 9.50599996940582e-06, "gpu_peak_gb": 89.37899398803711}
+{"step": 763500, "loss": 0.5598098039627075, "lr": 3e-05, "grad_norm": 0.2550642290735135, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111753967539, "tokens_per_second": 155473.45201527217, "data_wait_s": 9.86399982139119e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791254470.1e0c9aa3-e42e-441e-9e17-2e25895f438d-9bhv5.366.0.1791254470898546162-7724d9f6

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f2293166f3c66ca9857406ca64bf069813382735a2c5c857fe0543825b69cc0f
-size 72177
+oid sha256:6e8dd846362a08c296d324e010096143753ef637804eea33f8b1e0865374bd54
+size 74295