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": 78526.43525321044,
-      "required_pages_per_second_upper_estimate": 0.005432640997250033
+      "required_token_positions_per_second": 78096.44414136773,
+      "required_pages_per_second_upper_estimate": 0.005402893214415916
     },
     {
       "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": 9815.804406651305,
-      "required_pages_per_second_upper_estimate": 0.0041096723277601445
+      "required_token_positions_per_second": 9762.055517670966,
+      "required_pages_per_second_upper_estimate": 0.004087168790348511
     },
     {
       "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": 19631.60881330261,
-      "required_pages_per_second_upper_estimate": 0.004088746395471342
+      "required_token_positions_per_second": 19524.11103534193,
+      "required_pages_per_second_upper_estimate": 0.004066357443229177
     },
     {
       "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": 29447.413219953916,
-      "required_pages_per_second_upper_estimate": 0.001765327458939727
+      "required_token_positions_per_second": 29286.1665530129,
+      "required_pages_per_second_upper_estimate": 0.0017556609674660175
     },
     {
       "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": 29447.413219953916,
-      "required_pages_per_second_upper_estimate": 0.0022894068523693373
+      "required_token_positions_per_second": 29286.1665530129,
+      "required_pages_per_second_upper_estimate": 0.0022768706332638054
     },
     {
       "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": 29447.413219953916,
-      "required_pages_per_second_upper_estimate": 0.3777197986166662
+      "required_token_positions_per_second": 29286.1665530129,
+      "required_pages_per_second_upper_estimate": 0.3756514995063288
     }
   ],
   "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": 1.3701504740000132,
-    "producer_p95_seconds": 2.7244317560000013,
-    "gpu_update_seconds": 0.9388940140000273,
-    "producer_batches_per_second": 0.7298468445444557,
-    "gpu_batches_per_second": 1.0650829434300462,
-    "producer_headroom": 0.685248833479598,
-    "sustainable": false,
-    "buffer_seconds": 60.089216896001744,
-    "buffer_drain_seconds": 190.91022778499146,
+    "producer_batch_seconds": 0.04495540818750232,
+    "producer_p95_seconds": 0.11244288299997152,
+    "gpu_update_seconds": 0.9440634693500243,
+    "producer_batches_per_second": 22.244264712916156,
+    "gpu_batches_per_second": 1.059250815719506,
+    "producer_headroom": 20.999997718015948,
+    "sustainable": true,
+    "buffer_seconds": 60.420062038401554,
+    "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 6,
-    "max_pending_pages": 1490,
+    "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-v1/1791143623306220418-eb05a537/train.jsonl

@@ -0,0 +1 @@
+{"step": 685100, "loss": 0.5776553750038147, "lr": 8e-05, "grad_norm": 0.20740935451668885, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100281063608, "tokens_per_second": 152913.44909213492, "data_wait_s": 1.2569999967126932e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791143623.3eb6ddb3-ffce-42e5-bd48-c0d653189e98-8pb7d.369.0.1791143623306220418-eb05a537

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3c50ed21807fc54dc34cd657af6b2fc74bcd1d6117463b6c7c11329daa9f191a
-size 165
+oid sha256:daf7ee7f44d053951ecb7799891ef6a39930ced81d77766cbd2edc2d5449a20b
+size 2283