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": 78042.0667418929,
-      "required_pages_per_second_upper_estimate": 0.005399131259747297
+      "required_token_positions_per_second": 77503.68632549694,
+      "required_pages_per_second_upper_estimate": 0.0053618848533263505
     },
     {
       "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": 9755.258342736612,
-      "required_pages_per_second_upper_estimate": 0.0040843229551446505
+      "required_token_positions_per_second": 9687.960790687117,
+      "required_pages_per_second_upper_estimate": 0.004056146875434219
     },
     {
       "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": 19510.516685473223,
-      "required_pages_per_second_upper_estimate": 0.004063526098658639
+      "required_token_positions_per_second": 19375.921581374234,
+      "required_pages_per_second_upper_estimate": 0.004035493487986409
     },
     {
       "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": 29265.77502820983,
-      "required_pages_per_second_upper_estimate": 0.001754438526690129
+      "required_token_positions_per_second": 29063.882372061347,
+      "required_pages_per_second_upper_estimate": 0.001742335370226264
     },
     {
       "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": 29265.77502820983,
-      "required_pages_per_second_upper_estimate": 0.0022752852819032054
+      "required_token_positions_per_second": 29063.882372061347,
+      "required_pages_per_second_upper_estimate": 0.0022595890159196053
     },
     {
       "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": 29265.77502820983,
-      "required_pages_per_second_upper_estimate": 0.3753899389208685
+      "required_token_positions_per_second": 29063.882372061347,
+      "required_pages_per_second_upper_estimate": 0.3728002767032407
     }
   ],
   "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": 704566,
-    "producer_batch_seconds": 0.031540981624978315,
-    "producer_p95_seconds": 0.07767916900002092,
-    "gpu_update_seconds": 0.9447212647999095,
-    "producer_batches_per_second": 31.704783696651596,
-    "gpu_batches_per_second": 1.0585132750365247,
-    "producer_headroom": 29.952183354108243,
+    "producer_batch_seconds": 0.030661641929700778,
+    "producer_p95_seconds": 0.0766145359998518,
+    "gpu_update_seconds": 0.9512837839784813,
+    "producer_batches_per_second": 32.61403946640371,
+    "gpu_batches_per_second": 1.051211023295043,
+    "producer_headroom": 31.025206874424054,
     "sustainable": true,
-    "buffer_seconds": 60.46216094719421,
+    "buffer_seconds": 60.882162174622806,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 17,
+    "upload_pages_per_commit": 33,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790914434605211575-090606d4/train.jsonl

@@ -1 +1,2 @@
 {"step": 485100, "loss": 0.5988056063652039, "lr": 0.0003, "grad_norm": 0.17320443990552808, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 71002266879, "tokens_per_second": 154286.68957198376, "data_wait_s": 1.4927999927749624e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 485200, "loss": 0.5576065182685852, "lr": 0.0003, "grad_norm": 0.14968418877833817, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 71016926460, "tokens_per_second": 153832.48290073316, "data_wait_s": 1.3774999843008118e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1790914434.4dec8a9d-3312-413a-9119-9058bec1adb7-bx5st.8214.0.1790914434605211575-090606d4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0b524b838617d5663abe850f9e862b33ed0da5b0d8910f4e0b4d187fa06988ea
-size 2283
+oid sha256:92c6aaab3476e16b47ede0ed97d207db7137844f9c694891a16a3588f534b9f1
+size 4401