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": 76070.93164140428,
-      "required_pages_per_second_upper_estimate": 0.006593635554663342
+      "required_token_positions_per_second": 75749.71099442973,
+      "required_pages_per_second_upper_estimate": 0.006565792963109874
     },
     {
       "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": 9508.866455175536,
-      "required_pages_per_second_upper_estimate": 0.0037932289991924107
+      "required_token_positions_per_second": 9468.713874303716,
+      "required_pages_per_second_upper_estimate": 0.0037772115343480597
     },
     {
       "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": 19017.73291035107,
-      "required_pages_per_second_upper_estimate": 0.004399417808141945
+      "required_token_positions_per_second": 18937.42774860743,
+      "required_pages_per_second_upper_estimate": 0.00438084062229513
     },
     {
       "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": 28526.599365526607,
-      "required_pages_per_second_upper_estimate": 0.0017371202338606134
+      "required_token_positions_per_second": 28406.14162291115,
+      "required_pages_per_second_upper_estimate": 0.0017297849893282645
     },
     {
       "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": 28526.599365526607,
-      "required_pages_per_second_upper_estimate": 0.0021960565858393745
+      "required_token_positions_per_second": 28406.14162291115,
+      "required_pages_per_second_upper_estimate": 0.002186783415364463
     },
     {
       "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": 28526.599365526607,
-      "required_pages_per_second_upper_estimate": 0.35420489173332265
+      "required_token_positions_per_second": 28406.14162291115,
+      "required_pages_per_second_upper_estimate": 0.3527092097161696
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.04938664335156062,
-    "producer_p95_seconds": 0.13566265700001168,
-    "gpu_update_seconds": 0.969200697416869,
-    "producer_batches_per_second": 20.248389688715296,
-    "gpu_batches_per_second": 1.0317780441813733,
-    "producer_headroom": 19.624753407871406,
+    "producer_batch_seconds": 0.06370352673436486,
+    "producer_p95_seconds": 0.1352767229998335,
+    "gpu_update_seconds": 0.9733106441214225,
+    "producer_batches_per_second": 15.697718026976208,
+    "gpu_batches_per_second": 1.0274212103194136,
+    "producer_headroom": 15.27875604407268,
     "sustainable": true,
-    "buffer_seconds": 60.090443239845875,
+    "buffer_seconds": 60.3452599355282,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 67,
+    "upload_pages_per_commit": 62,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790563279934888633-82973089/train.jsonl

@@ -9,3 +9,4 @@
 {"step": 155900, "loss": 0.5951595827937126, "lr": 0.0003, "grad_norm": 0.15149574986578188, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22815968278, "tokens_per_second": 150860.53562940232, "data_wait_s": 1.3899000123274163e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 156000, "loss": 0.6793905571103096, "lr": 0.0003, "grad_norm": 0.1406019373286998, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22830648258, "tokens_per_second": 150100.27176993177, "data_wait_s": 1.3684000123248552e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 156100, "loss": 0.5846447832882404, "lr": 0.0003, "grad_norm": 0.13303587614305679, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22845328580, "tokens_per_second": 151130.18552196387, "data_wait_s": 1.2695999885181664e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 156200, "loss": 0.576857328414917, "lr": 0.0003, "grad_norm": 0.13306192544028528, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22859969626, "tokens_per_second": 150264.58248178678, "data_wait_s": 1.6371000128856394e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790563279.bdfa290c-aca9-4e96-a5d2-8425921a3278-d85hq.1304.0.1790563279934888633-82973089

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5c771cd9ee17fe197904301e9a06f45c56e1402233a26a23f411a6f13d585272
-size 23463
+oid sha256:8313e75dee4ddb2634c36a1ee4f6b41f9f85cf464bf0c838742cb491e3b6df52
+size 25581