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": 75706.35071992404,
-      "required_pages_per_second_upper_estimate": 0.006562034604411332
+      "required_token_positions_per_second": 76025.74682254261,
+      "required_pages_per_second_upper_estimate": 0.006589719049084289
     },
     {
       "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": 9463.293839990505,
-      "required_pages_per_second_upper_estimate": 0.0037750494016237854
+      "required_token_positions_per_second": 9503.218352817827,
+      "required_pages_per_second_upper_estimate": 0.0037909758867152653
     },
     {
       "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": 18926.58767998101,
-      "required_pages_per_second_upper_estimate": 0.0043783329631971
+      "required_token_positions_per_second": 19006.436705635653,
+      "required_pages_per_second_upper_estimate": 0.004396804629987447
     },
     {
       "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": 28389.881519971514,
-      "required_pages_per_second_upper_estimate": 0.0017287948343694076
+      "required_token_positions_per_second": 28509.655058453478,
+      "required_pages_per_second_upper_estimate": 0.0017360884144597684
     },
     {
       "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": 28389.881519971514,
-      "required_pages_per_second_upper_estimate": 0.0021855316676293243
+      "required_token_positions_per_second": 28509.655058453478,
+      "required_pages_per_second_upper_estimate": 0.002194752166176041
     },
     {
       "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": 28389.881519971514,
-      "required_pages_per_second_upper_estimate": 0.35250731365672316
+      "required_token_positions_per_second": 28509.655058453478,
+      "required_pages_per_second_upper_estimate": 0.3539945001484222
     }
   ],
   "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": 0.05480057218749046,
-    "producer_p95_seconds": 0.14381412100010493,
-    "gpu_update_seconds": 0.9738681008777856,
-    "producer_batches_per_second": 18.24798464838427,
-    "gpu_batches_per_second": 1.026833098957303,
-    "producer_headroom": 17.77113015436898,
+    "producer_batch_seconds": 0.050725168539052845,
+    "producer_p95_seconds": 0.14474023299999317,
+    "gpu_update_seconds": 0.969776728035228,
+    "producer_batches_per_second": 19.71407939689957,
+    "gpu_batches_per_second": 1.0311651858526287,
+    "producer_headroom": 19.118255413751967,
     "sustainable": true,
-    "buffer_seconds": 60.379822254422706,
+    "buffer_seconds": 60.12615713818413,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 67,
+    "upload_pages_per_commit": 71,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -14,3 +14,4 @@
 {"step": 156400, "loss": 0.5991663802415133, "lr": 0.0003, "grad_norm": 0.14258397599734335, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22889448533, "tokens_per_second": 151256.05648356315, "data_wait_s": 1.3411000054475153e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 156500, "loss": 0.6319344639778137, "lr": 0.0003, "grad_norm": 0.14753616763741959, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22904209759, "tokens_per_second": 151507.8120863896, "data_wait_s": 1.3636999938171357e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 156600, "loss": 0.6385067477822304, "lr": 0.0003, "grad_norm": 0.15079786627480568, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22918824594, "tokens_per_second": 150448.6576303968, "data_wait_s": 1.2602999959199224e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 156700, "loss": 0.6273212861269712, "lr": 0.0003, "grad_norm": 0.14089745148802876, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22933594966, "tokens_per_second": 151905.57121771606, "data_wait_s": 1.25619999380433e-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:b96e76d8719f37d4a8dff03fd73e4ad13d79d2044514d255c3cdaa375403a0f9
-size 34053
+oid sha256:636047af403283a6ddc680818f6f5718bdffc3d6c4852f90c8c78500c43a6cd6
+size 36171