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": 76367.5904023782,
-      "required_pages_per_second_upper_estimate": 0.006619349184189806
+      "required_token_positions_per_second": 76195.63474790327,
+      "required_pages_per_second_upper_estimate": 0.00660444450387754
     },
     {
       "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": 9545.948800297276,
-      "required_pages_per_second_upper_estimate": 0.00380802170109194
+      "required_token_positions_per_second": 9524.454343487909,
+      "required_pages_per_second_upper_estimate": 0.0037994472408999155
     },
     {
       "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": 19091.89760059455,
-      "required_pages_per_second_upper_estimate": 0.004416574503975806
+      "required_token_positions_per_second": 19048.908686975818,
+      "required_pages_per_second_upper_estimate": 0.004406629775389152
     },
     {
       "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": 28637.846400891827,
-      "required_pages_per_second_upper_estimate": 0.0017438945946462688
+      "required_token_positions_per_second": 28573.36303046373,
+      "required_pages_per_second_upper_estimate": 0.0017399678957052958
     },
     {
       "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": 28637.846400891827,
-      "required_pages_per_second_upper_estimate": 0.0022046206905732926
+      "required_token_positions_per_second": 28573.36303046373,
+      "required_pages_per_second_upper_estimate": 0.002199656582216344
     },
     {
       "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": 28637.846400891827,
-      "required_pages_per_second_upper_estimate": 0.35558620759268195
+      "required_token_positions_per_second": 28573.36303046373,
+      "required_pages_per_second_upper_estimate": 0.35478553994392303
     }
   ],
   "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": 0.04622200666405263,
-    "producer_p95_seconds": 0.12276615500013577,
-    "gpu_update_seconds": 0.9654357249132741,
-    "producer_batches_per_second": 21.63471627850617,
-    "gpu_batches_per_second": 1.0358017361433676,
-    "producer_headroom": 20.886927993632614,
+    "producer_batch_seconds": 0.04657719053124865,
+    "producer_p95_seconds": 0.13263870699984182,
+    "gpu_update_seconds": 0.9676144866294827,
+    "producer_batches_per_second": 21.46973633648212,
+    "gpu_batches_per_second": 1.0334694383124903,
+    "producer_headroom": 20.774427903295496,
     "sustainable": true,
-    "buffer_seconds": 60.822450669536266,
+    "buffer_seconds": 60.959712657657406,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 72,
+    "upload_pages_per_commit": 75,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -15,3 +15,4 @@
 {"step": 41500, "loss": 0.6294730324298143, "lr": 0.0003, "grad_norm": 0.13219610104495547, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6072666804, "tokens_per_second": 151918.62530588364, "data_wait_s": 9.390999821334844e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 41600, "loss": 0.6368809249252081, "lr": 0.0003, "grad_norm": 0.15239497157121357, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6087361539, "tokens_per_second": 151682.4203449486, "data_wait_s": 9.697000223241048e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 41700, "loss": 0.645052120089531, "lr": 0.0003, "grad_norm": 0.13222859382635524, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6102058721, "tokens_per_second": 151438.91375842685, "data_wait_s": 9.574000159773277e-06, "gpu_peak_gb": 85.46427917480469}
+{"step": 41800, "loss": 0.6425942555069923, "lr": 0.0003, "grad_norm": 0.13991135484243736, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 6116836242, "tokens_per_second": 152251.9526339524, "data_wait_s": 1.2988999969820725e-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 sha256:736559bd4cfe5a6a67f4e7abf371c2cbab47f2f6bddd9c622a04ee223bbfb1c7
-size 36171
+oid sha256:6f57732c12fc1019cc22202a989c4583e5308eeadc8ff78e9f2e90972762207b
+size 38289