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": 78036.6325342031,
-      "required_pages_per_second_upper_estimate": 0.005398755308650219
+      "required_token_positions_per_second": 77618.29944781002,
+      "required_pages_per_second_upper_estimate": 0.005369814055067052
     },
     {
       "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": 9754.579066775388,
-      "required_pages_per_second_upper_estimate": 0.0040840385564845805
+      "required_token_positions_per_second": 9702.287430976252,
+      "required_pages_per_second_upper_estimate": 0.0040621451405490104
     },
     {
       "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": 19509.158133550776,
-      "required_pages_per_second_upper_estimate": 0.004063243148120618
+      "required_token_positions_per_second": 19404.574861952504,
+      "required_pages_per_second_upper_estimate": 0.00404146121069304
     },
     {
       "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": 29263.737200326166,
-      "required_pages_per_second_upper_estimate": 0.0017543163620200866
+      "required_token_positions_per_second": 29106.862292928756,
+      "required_pages_per_second_upper_estimate": 0.0017449119508557258
     },
     {
       "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": 29263.737200326166,
-      "required_pages_per_second_upper_estimate": 0.002275126849748708
+      "required_token_positions_per_second": 29106.862292928756,
+      "required_pages_per_second_upper_estimate": 0.0022629305157183537
     },
     {
       "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": 29263.737200326166,
-      "required_pages_per_second_upper_estimate": 0.375363799852826
+      "required_token_positions_per_second": 29106.862292928756,
+      "required_pages_per_second_upper_estimate": 0.37335157697988425
     }
   ],
   "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": 0.5398314225000149,
-    "producer_p95_seconds": 1.0657202160000452,
-    "gpu_update_seconds": 0.9447870520000379,
-    "producer_batches_per_second": 1.852430144523446,
-    "gpu_batches_per_second": 1.0584395688775379,
-    "producer_headroom": 1.7501520152803105,
+    "producer_batch_seconds": 0.04679934131247876,
+    "producer_p95_seconds": 0.10147988700009591,
+    "gpu_update_seconds": 0.9498790945500446,
+    "producer_batches_per_second": 21.367822109354265,
+    "gpu_batches_per_second": 1.0527655632569717,
+    "producer_headroom": 20.296847517739852,
     "sustainable": true,
-    "buffer_seconds": 60.466371328002424,
+    "buffer_seconds": 60.792262051202854,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 7,
-    "max_pending_pages": 1681,
+    "upload_pages_per_commit": 17,
+    "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,
     "upload_sustainable": null,

metrics/void-byte-v1/1790956870018829698-ff939fa5/train.jsonl

@@ -0,0 +1 @@
+{"step": 525100, "loss": 0.6340075135231018, "lr": 0.0003, "grad_norm": 0.16481391997628744, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 76857132722, "tokens_per_second": 154353.8726525625, "data_wait_s": 6.23400001131813e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1790956870.abdf5fd7-18c6-4e42-b51e-c994b21d5ba1-dh6p5.3284.0.1790956870018829698-ff939fa5

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a488995a361c029ba8a0dc405a1966498ee3ae4622824908dd932508c06acf51
-size 165
+oid sha256:e4d7b5223e20ce33a36db24e5f9f292063f58d33b493d035806dee716b141e8d
+size 2283