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": 78098.01702812791,
-      "required_pages_per_second_upper_estimate": 0.005403002030371586
+      "required_token_positions_per_second": 77679.03479490291,
+      "required_pages_per_second_upper_estimate": 0.00537401586730436
     },
     {
       "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": 9762.25212851599,
-      "required_pages_per_second_upper_estimate": 0.004087251107203621
+      "required_token_positions_per_second": 9709.879349362864,
+      "required_pages_per_second_upper_estimate": 0.004065323718240201
     },
     {
       "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": 19524.50425703198,
-      "required_pages_per_second_upper_estimate": 0.00406643934093726
+      "required_token_positions_per_second": 19419.758698725727,
+      "required_pages_per_second_upper_estimate": 0.004044623603468199
     },
     {
       "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": 29286.756385547968,
-      "required_pages_per_second_upper_estimate": 0.0017556963270258727
+      "required_token_positions_per_second": 29129.638048088593,
+      "required_pages_per_second_upper_estimate": 0.0017462773226009935
     },
     {
       "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": 29286.756385547968,
-      "required_pages_per_second_upper_estimate": 0.0022769164901490077
+      "required_token_positions_per_second": 29129.638048088593,
+      "required_pages_per_second_upper_estimate": 0.002264701230502071
     },
     {
       "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": 29286.756385547968,
-      "required_pages_per_second_upper_estimate": 0.3756590652447758
+      "required_token_positions_per_second": 29129.638048088593,
+      "required_pages_per_second_upper_estimate": 0.37364371991237405
     }
   ],
   "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": 1.7146579959999997,
-    "producer_p95_seconds": 3.415684295999995,
-    "gpu_update_seconds": 0.9440444560000287,
-    "producer_batches_per_second": 0.5832066816431188,
-    "gpu_batches_per_second": 1.059272149361544,
-    "producer_headroom": 0.5505730345073601,
-    "sustainable": false,
-    "buffer_seconds": 60.41884518400184,
-    "buffer_drain_seconds": 134.43529165583922,
+    "producer_batch_seconds": 0.04979349493750007,
+    "producer_p95_seconds": 0.11288187099995639,
+    "gpu_update_seconds": 0.9491364072000266,
+    "producer_batches_per_second": 20.0829445945737,
+    "gpu_batches_per_second": 1.0535893391235747,
+    "producer_headroom": 19.061453878490877,
+    "sustainable": true,
+    "buffer_seconds": 60.7447300608017,
+    "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 8,
-    "max_pending_pages": 1806,
+    "upload_pages_per_commit": 19,
+    "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,
     "upload_sustainable": null,

metrics/void-byte-v1/1791463982175382982-d4a3b372/train.jsonl

@@ -0,0 +1 @@
+{"step": 925100, "loss": 0.5094915628433228, "lr": 3e-05, "grad_norm": 0.3323485047186109, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135408062366, "tokens_per_second": 152551.99412574965, "data_wait_s": 7.469000024684647e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791463982.2e076f0c-3fe2-4316-8f17-4e30b2e706e2-b5972.360.0.1791463982175382982-d4a3b372

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b65fc44666df32f089915a1a00aec1b2141ee82423f64dea5713a3d8a396621e
-size 165
+oid sha256:82dce1d51a1f681705d7401b210b4d8e4167a9c3f8738010e15c909b2ffaeb60
+size 2283