Update TensorBoard training metrics

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 1062967.6168812835,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 76307.27473876862,
-      "required_pages_per_second_upper_estimate": 0.006614121175336779
+      "required_token_positions_per_second": 76344.00351364,
+      "required_pages_per_second_upper_estimate": 0.0066173047324544
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 1206672.263377472,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9538.409342346078,
-      "required_pages_per_second_upper_estimate": 0.0038050140985902655
+      "required_token_positions_per_second": 9543.000439205,
+      "required_pages_per_second_upper_estimate": 0.0038068455557703045
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 410042.7581134458,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19076.818684692156,
-      "required_pages_per_second_upper_estimate": 0.004413086262161643
+      "required_token_positions_per_second": 19086.00087841,
+      "required_pages_per_second_upper_estimate": 0.004415210401077176
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 2163486.5624068812,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28615.228027038233,
-      "required_pages_per_second_upper_estimate": 0.0017425172543480214
+      "required_token_positions_per_second": 28629.001317614995,
+      "required_pages_per_second_upper_estimate": 0.0017433559754812752
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 1546694.2737633225,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28615.228027038233,
-      "required_pages_per_second_upper_estimate": 0.002202879465542377
+      "required_token_positions_per_second": 28629.001317614995,
+      "required_pages_per_second_upper_estimate": 0.0022039397715778847
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 22,
-      "required_token_positions_per_second": 28615.228027038233,
-      "required_pages_per_second_upper_estimate": 0.35530536308824806
+      "required_token_positions_per_second": 28629.001317614995,
+      "required_pages_per_second_upper_estimate": 0.35547638126097314
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.04953015893748969,
-    "producer_p95_seconds": 0.13143097700003636,
-    "gpu_update_seconds": 0.9661988355946592,
-    "producer_batches_per_second": 20.18971918224744,
-    "gpu_batches_per_second": 1.0349836525983158,
-    "producer_headroom": 19.507283164870632,
+    "producer_batch_seconds": 0.05163340690622675,
+    "producer_p95_seconds": 0.1392136389999905,
+    "gpu_update_seconds": 0.9657340014507806,
+    "producer_batches_per_second": 19.367306167035135,
+    "gpu_batches_per_second": 1.0354818184901258,
+    "producer_headroom": 18.70366608201322,
     "sustainable": true,
-    "buffer_seconds": 60.87052664246353,
+    "buffer_seconds": 60.841242091399174,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 84,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790434090210070261-a05fc3fb/train.jsonl

@@ -22,3 +22,4 @@
 {"step": 42200, "loss": 0.6160119939595461, "lr": 0.0003, "grad_norm": 0.12473146730710874, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6175109717, "tokens_per_second": 150601.47482535624, "data_wait_s": 1.2016999789921101e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 42300, "loss": 0.6377940252423286, "lr": 0.0003, "grad_norm": 0.12067368401402269, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6189763884, "tokens_per_second": 151108.72302846916, "data_wait_s": 1.1086000085924752e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 42400, "loss": 0.6546931639313698, "lr": 0.0003, "grad_norm": 0.13572722711902627, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6204361023, "tokens_per_second": 150572.9859644851, "data_wait_s": 9.984999906009762e-06, "gpu_peak_gb": 85.46427917480469}
+{"step": 42500, "loss": 0.6561312675476074, "lr": 0.0003, "grad_norm": 0.13010470941836758, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6218832513, "tokens_per_second": 148977.37935472347, "data_wait_s": 1.1410999832150992e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790434090.372bd512-57cc-4928-b44a-3b375a0fa071-5spg5.451.0.1790434090210070261-a05fc3fb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8cff39f79f57d370877288e9c1a075207bac5241c07271fcc1549807a4d0530b
-size 50997
+oid sha256:3c89bba23577bc7d4da46230cc070784a56ad74b17714d1f35e5f882a21c581f
+size 53115