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": 77678.31331671114,
-      "required_pages_per_second_upper_estimate": 0.005373965953768993
+      "required_token_positions_per_second": 77744.9244637589,
+      "required_pages_per_second_upper_estimate": 0.005378574267480913
     },
     {
       "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": 9709.789164588892,
-      "required_pages_per_second_upper_estimate": 0.004065285959758611
+      "required_token_positions_per_second": 9718.115557969862,
+      "required_pages_per_second_upper_estimate": 0.004068772046792358
     },
     {
       "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": 19419.578329177784,
-      "required_pages_per_second_upper_estimate": 0.004044586037248027
+      "required_token_positions_per_second": 19436.231115939725,
+      "required_pages_per_second_upper_estimate": 0.00404805437356702
     },
     {
       "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": 29129.367493766673,
-      "required_pages_per_second_upper_estimate": 0.001746261103282507
+      "required_token_positions_per_second": 29154.346673909586,
+      "required_pages_per_second_upper_estimate": 0.001747758566991075
     },
     {
       "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": 29129.367493766673,
-      "required_pages_per_second_upper_estimate": 0.0022646801960935848
+      "required_token_positions_per_second": 29154.346673909586,
+      "required_pages_per_second_upper_estimate": 0.0022666222174778945
     },
     {
       "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": 29129.367493766673,
-      "required_pages_per_second_upper_estimate": 0.37364024953203107
+      "required_token_positions_per_second": 29154.346673909586,
+      "required_pages_per_second_upper_estimate": 0.3739606556343503
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.04750775897656645,
-    "producer_p95_seconds": 0.11337590900006944,
-    "gpu_update_seconds": 0.9491452228036305,
-    "producer_batches_per_second": 21.049193259005488,
-    "gpu_batches_per_second": 1.0535795534493155,
-    "producer_headroom": 19.97874122565544,
+    "producer_batch_seconds": 0.054468098054687886,
+    "producer_p95_seconds": 0.12858819400003085,
+    "gpu_update_seconds": 0.9483320037743248,
+    "producer_batches_per_second": 18.359370635559273,
+    "gpu_batches_per_second": 1.054483024953327,
+    "producer_headroom": 17.410778742855427,
     "sustainable": true,
-    "buffer_seconds": 60.74529425943235,
+    "buffer_seconds": 60.69324824155679,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 60,
+    "upload_pages_per_commit": 67,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791399463488568880-054ba646/train.jsonl

@@ -4,3 +4,4 @@
 {"step": 885400, "loss": 0.5317869782447815, "lr": 3e-05, "grad_norm": 0.3113006330539896, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 129597709606, "tokens_per_second": 154322.61474139697, "data_wait_s": 1.4046000046619156e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 885500, "loss": 0.514037013053894, "lr": 3e-05, "grad_norm": 0.32076586136130014, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 129612357157, "tokens_per_second": 154267.41349759378, "data_wait_s": 1.0085999974762672e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 885600, "loss": 0.5332300066947937, "lr": 3e-05, "grad_norm": 0.3074175699791647, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 129627117269, "tokens_per_second": 155506.79471233697, "data_wait_s": 9.958999953596503e-06, "gpu_peak_gb": 89.37899398803711}
+{"step": 885700, "loss": 0.5637478828430176, "lr": 3e-05, "grad_norm": 0.3168497868655922, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 129641887771, "tokens_per_second": 155400.9375584298, "data_wait_s": 8.966999985204893e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791399463.60f38c85-66c9-4fe6-8e58-7b3a56dc2cee-wtkw8.346.0.1791399463488568880-054ba646

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1a7d9cec36229d8c150aa656919ba649ace5cf693d3890ded04ecdff85c93ee0
-size 12873
+oid sha256:2be8beb3034998e16f162aedcc895928d7be99e281096832d0f9e54c21ccef6c
+size 14991