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": 78567.9080594358,
-      "required_pages_per_second_upper_estimate": 0.005435510182214872
+      "required_token_positions_per_second": 78125.88580543191,
+      "required_pages_per_second_upper_estimate": 0.0054049300570499495
     },
     {
       "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": 9820.988507429474,
-      "required_pages_per_second_upper_estimate": 0.004111842802499629
+      "required_token_positions_per_second": 9765.73572567899,
+      "required_pages_per_second_upper_estimate": 0.004088709616589988
     },
     {
       "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": 19641.97701485895,
-      "required_pages_per_second_upper_estimate": 0.004090905818427665
+      "required_token_positions_per_second": 19531.47145135798,
+      "required_pages_per_second_upper_estimate": 0.004067890423778096
     },
     {
       "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": 29462.96552228842,
-      "required_pages_per_second_upper_estimate": 0.00176625979571769
+      "required_token_positions_per_second": 29297.207177036966,
+      "required_pages_per_second_upper_estimate": 0.0017563228360182035
     },
     {
       "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": 29462.96552228842,
-      "required_pages_per_second_upper_estimate": 0.0022906159754684965
+      "required_token_positions_per_second": 29297.207177036966,
+      "required_pages_per_second_upper_estimate": 0.002277728993219104
     },
     {
       "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": 29462.96552228842,
-      "required_pages_per_second_upper_estimate": 0.377919286852252
+      "required_token_positions_per_second": 29297.207177036966,
+      "required_pages_per_second_upper_estimate": 0.37579311677681104
     }
   ],
   "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": 1.235392456999989,
-    "producer_p95_seconds": 2.3715912330000037,
-    "gpu_update_seconds": 0.9383984099999907,
-    "producer_batches_per_second": 0.8094593700437406,
-    "gpu_batches_per_second": 1.0656454543651772,
-    "producer_headroom": 0.7595953858086403,
-    "sustainable": false,
-    "buffer_seconds": 60.057498239999404,
-    "buffer_drain_seconds": 249.81840902685101,
+    "producer_batch_seconds": 0.04667298129687936,
+    "producer_p95_seconds": 0.12698847199999364,
+    "gpu_update_seconds": 0.9437076999499936,
+    "producer_batches_per_second": 21.42567224577235,
+    "gpu_batches_per_second": 1.0596501438453765,
+    "producer_headroom": 20.21957187494024,
+    "sustainable": true,
+    "buffer_seconds": 60.39729279679959,
+    "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 7,
-    "max_pending_pages": 1737,
+    "upload_pages_per_commit": 22,
+    "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,
     "upload_sustainable": null,

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

@@ -0,0 +1 @@
+{"step": 440300, "loss": 0.6032825112342834, "lr": 0.0003, "grad_norm": 0.1660740186333239, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 64447028234, "tokens_per_second": 152800.35349740396, "data_wait_s": 1.984599998650083e-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 sha256:3da923b2fc140153b6ed401074dc15860e89c403d2e7f9b8695e6c52db1890e0
-size 165
+oid sha256:c6b7530e7f492e49c3b991107f1500fcc971e933950a07e5e76cf2c29b8ce342
+size 2283