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": 75806.73498936085,
-      "required_pages_per_second_upper_estimate": 0.00657073565846451
+      "required_token_positions_per_second": 75951.9173464346,
+      "required_pages_per_second_upper_estimate": 0.006583319697214342
     },
     {
       "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": 9475.841873670106,
-      "required_pages_per_second_upper_estimate": 0.0037800549998684005
+      "required_token_positions_per_second": 9493.989668304324,
+      "required_pages_per_second_upper_estimate": 0.0037872944264816994
     },
     {
       "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": 18951.683747340212,
-      "required_pages_per_second_upper_estimate": 0.004384138496704951
+      "required_token_positions_per_second": 18987.97933660865,
+      "required_pages_per_second_upper_estimate": 0.004392534842501649
     },
     {
       "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": 28427.525621010318,
-      "required_pages_per_second_upper_estimate": 0.0017310871626193365
+      "required_token_positions_per_second": 28481.969004912975,
+      "required_pages_per_second_upper_estimate": 0.0017344024790566486
     },
     {
       "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": 28427.525621010318,
-      "required_pages_per_second_upper_estimate": 0.0021884296147328328
+      "required_token_positions_per_second": 28481.969004912975,
+      "required_pages_per_second_upper_estimate": 0.0021926208171345915
     },
     {
       "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": 28427.525621010318,
-      "required_pages_per_second_upper_estimate": 0.3529747274049234
+      "required_token_positions_per_second": 28481.969004912975,
+      "required_pages_per_second_upper_estimate": 0.35365073202270975
     }
   ],
   "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": 717798,
-    "producer_batch_seconds": 0.04272710308592487,
-    "producer_p95_seconds": 0.12297350799985907,
-    "gpu_update_seconds": 0.9725784920079649,
-    "producer_batches_per_second": 23.404348242121273,
-    "gpu_batches_per_second": 1.0281946477506625,
-    "producer_headroom": 22.76256571975157,
+    "producer_batch_seconds": 0.05116541806249586,
+    "producer_p95_seconds": 0.1363660549998258,
+    "gpu_update_seconds": 0.9707194048006611,
+    "producer_batches_per_second": 19.544450878492828,
+    "gpu_batches_per_second": 1.0301638094948269,
+    "producer_headroom": 18.972177723926315,
     "sustainable": true,
-    "buffer_seconds": 60.29986650449382,
+    "buffer_seconds": 60.18460309764099,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 68,
+    "upload_pages_per_commit": 67,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790782192342215557-281f6532/train.jsonl

@@ -11,3 +11,4 @@
 {"step": 356100, "loss": 0.5715948585420847, "lr": 0.0003, "grad_norm": 0.14480262907768615, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52122825287, "tokens_per_second": 151006.9257437939, "data_wait_s": 7.612999979755841e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 356200, "loss": 0.5747460946440697, "lr": 0.0003, "grad_norm": 0.13920962199559656, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52137511384, "tokens_per_second": 150696.00068849837, "data_wait_s": 1.3461000207826146e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 356300, "loss": 0.5986430812627077, "lr": 0.0003, "grad_norm": 0.15324332866862694, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52152238141, "tokens_per_second": 151452.92743448433, "data_wait_s": 1.3626000054500764e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 356400, "loss": 0.6035982519388199, "lr": 0.0003, "grad_norm": 0.14425823427770984, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52166934931, "tokens_per_second": 151385.60718366815, "data_wait_s": 1.2531000038507045e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790782192.062ccfe6-664f-43fe-9e0e-62f43f0c4870-64tpq.429.0.1790782192342215557-281f6532

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4759b8c0c447a15c7df1a57485bcd9c7ef31627306658cab600d70e074b01860
-size 27699
+oid sha256:8255bf0a7534bf05e3a58172c644f93d29834a83ced3381afa4d50384bb3a136
+size 29817