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": 78404.75145554924,
-      "required_pages_per_second_upper_estimate": 0.0054242226297825215
+      "required_token_positions_per_second": 77982.72443368332,
+      "required_pages_per_second_upper_estimate": 0.005395025821172234
     },
     {
       "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": 9800.593931943655,
-      "required_pages_per_second_upper_estimate": 0.004103304019630882
+      "required_token_positions_per_second": 9747.840554210416,
+      "required_pages_per_second_upper_estimate": 0.004081217281989771
     },
     {
       "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": 19601.18786388731,
-      "required_pages_per_second_upper_estimate": 0.00408241051396252
+      "required_token_positions_per_second": 19495.68110842083,
+      "required_pages_per_second_upper_estimate": 0.004060436239199103
     },
     {
       "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": 29401.78179583096,
-      "required_pages_per_second_upper_estimate": 0.0017625919247387062
+      "required_token_positions_per_second": 29243.521662631247,
+      "required_pages_per_second_upper_estimate": 0.0017531044714026148
     },
     {
       "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": 29401.78179583096,
-      "required_pages_per_second_upper_estimate": 0.0022858592098552004
+      "required_token_positions_per_second": 29243.521662631247,
+      "required_pages_per_second_upper_estimate": 0.002273555180611681
     },
     {
       "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": 29401.78179583096,
-      "required_pages_per_second_upper_estimate": 0.3771344877032229
+      "required_token_positions_per_second": 29243.521662631247,
+      "required_pages_per_second_upper_estimate": 0.3751044966410288
     }
   ],
   "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": 720750,
-    "producer_batch_seconds": 1.462310465999991,
-    "producer_p95_seconds": 2.9105605440000204,
-    "gpu_update_seconds": 0.9403511730000105,
-    "producer_batches_per_second": 0.6838493078254465,
-    "gpu_batches_per_second": 1.0634325012959693,
-    "producer_headroom": 0.6430584987689039,
-    "sustainable": false,
-    "buffer_seconds": 60.18247507200067,
-    "buffer_drain_seconds": 168.6059896773855,
+    "producer_batch_seconds": 0.05104165732812849,
+    "producer_p95_seconds": 0.14827627099998608,
+    "gpu_update_seconds": 0.9454401668500111,
+    "producer_batches_per_second": 19.59184031919965,
+    "gpu_batches_per_second": 1.0577083934690121,
+    "producer_headroom": 18.52291278028289,
+    "sustainable": true,
+    "buffer_seconds": 60.50817067840071,
+    "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 7,
-    "max_pending_pages": 1646,
+    "upload_pages_per_commit": 18,
+    "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,
     "upload_sustainable": null,

metrics/void-byte-v1/1791083219122319327-61ae60ac/train.jsonl

@@ -0,0 +1 @@
+{"step": 645100, "loss": 0.5672259330749512, "lr": 0.0001, "grad_norm": 0.19234217093846068, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94425964227, "tokens_per_second": 153596.82575021993, "data_wait_s": 1.702800000202842e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791083219.1a73f132-ede3-4b51-9e4f-0cafff989449-8p4r7.435.0.1791083219122319327-61ae60ac

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:12f4094f6e83c188ae95add91457868cc3e58d99e7ec440ecf6a5cac646b0976
-size 165
+oid sha256:85a2bf2ceef22509ca9d8d3221edf5c2f39b816ba93fda28e59f6179b678d304
+size 2283