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": 78631.56596754376,
-      "required_pages_per_second_upper_estimate": 0.005439914184004478
+      "required_token_positions_per_second": 78182.84309159202,
+      "required_pages_per_second_upper_estimate": 0.005408870494265617
     },
     {
       "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": 9828.94574594297,
-      "required_pages_per_second_upper_estimate": 0.004115174332099195
+      "required_token_positions_per_second": 9772.855386449002,
+      "required_pages_per_second_upper_estimate": 0.004091690469878969
     },
     {
       "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": 19657.89149188594,
-      "required_pages_per_second_upper_estimate": 0.004094220384299405
+      "required_token_positions_per_second": 19545.710772898005,
+      "required_pages_per_second_upper_estimate": 0.004070856098938722
     },
     {
       "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": 29486.83723782891,
-      "required_pages_per_second_upper_estimate": 0.0017676908686143431
+      "required_token_positions_per_second": 29318.566159347007,
+      "required_pages_per_second_upper_estimate": 0.001757603274394413
     },
     {
       "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": 29486.83723782891,
-      "required_pages_per_second_upper_estimate": 0.00229247189635119
+      "required_token_positions_per_second": 29318.566159347007,
+      "required_pages_per_second_upper_estimate": 0.002279389560145475
     },
     {
       "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": 29486.83723782891,
-      "required_pages_per_second_upper_estimate": 0.37822548758775426
+      "required_token_positions_per_second": 29318.566159347007,
+      "required_pages_per_second_upper_estimate": 0.37606708686839585
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,18 +158,18 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.9959551044994441,
-    "producer_p95_seconds": 1.9783748859990737,
-    "gpu_update_seconds": 0.9376387089942,
-    "producer_batches_per_second": 1.0040613231282034,
-    "gpu_batches_per_second": 1.0665088700024923,
-    "producer_headroom": 0.9414467627689369,
-    "sustainable": false,
-    "buffer_seconds": 60.0088773756288,
-    "buffer_drain_seconds": 1024.8601138622196,
+    "producer_batch_seconds": 0.07368372543703572,
+    "producer_p95_seconds": 0.24560894800379174,
+    "gpu_update_seconds": 0.9430201957944517,
+    "producer_batches_per_second": 13.571517917542332,
+    "gpu_batches_per_second": 1.0604226764810114,
+    "producer_headroom": 12.798215483828681,
+    "sustainable": true,
+    "buffer_seconds": 60.35329253084491,
+    "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 6,
-    "max_pending_pages": 1509,
+    "upload_pages_per_commit": 21,
+    "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,
     "upload_sustainable": null,

metrics/void-byte-v1/1790911149483311115-491a3c82/train.jsonl

@@ -0,0 +1 @@
+{"step": 483300, "loss": 0.566653311252594, "lr": 0.0003, "grad_norm": 0.14878479019944973, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 70738479781, "tokens_per_second": 151164.7510008121, "data_wait_s": 9.629999112803489e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1790911149.f92f6aaf-743a-4c24-a54a-850bd75954cb-8js8b.42163.0.1790911149483311115-491a3c82

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:78231dd24753b42834d55849188c1ea9ef651dd543f472863ea43c3a21925071
-size 165
+oid sha256:5e39d5e616bec41244a4cc9ae60f73c7cb3d015978055bc009b5875f033b56d6
+size 2283