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": 76025.74682254261,
-      "required_pages_per_second_upper_estimate": 0.006589719049084289
+      "required_token_positions_per_second": 76081.06344547702,
+      "required_pages_per_second_upper_estimate": 0.006594513753761543
     },
     {
       "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": 9503.218352817827,
-      "required_pages_per_second_upper_estimate": 0.0037909758867152653
+      "required_token_positions_per_second": 9510.132930684627,
+      "required_pages_per_second_upper_estimate": 0.003793734215208484
     },
     {
       "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": 19006.436705635653,
-      "required_pages_per_second_upper_estimate": 0.004396804629987447
+      "required_token_positions_per_second": 19020.265861369255,
+      "required_pages_per_second_upper_estimate": 0.004400003761781587
     },
     {
       "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": 28509.655058453478,
-      "required_pages_per_second_upper_estimate": 0.0017360884144597684
+      "required_token_positions_per_second": 28530.398792053882,
+      "required_pages_per_second_upper_estimate": 0.001737351599002074
     },
     {
       "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": 28509.655058453478,
-      "required_pages_per_second_upper_estimate": 0.002194752166176041
+      "required_token_positions_per_second": 28530.398792053882,
+      "required_pages_per_second_upper_estimate": 0.0021963490762109297
     },
     {
       "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": 28509.655058453478,
-      "required_pages_per_second_upper_estimate": 0.3539945001484222
+      "required_token_positions_per_second": 28530.398792053882,
+      "required_pages_per_second_upper_estimate": 0.3542520678949288
     }
   ],
   "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": 0.050725168539052845,
-    "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,
+    "producer_batch_seconds": 0.05477003137500702,
+    "producer_p95_seconds": 0.15514690700001665,
+    "gpu_update_seconds": 0.969071627828082,
+    "producer_batches_per_second": 18.25816007212159,
+    "gpu_batches_per_second": 1.0319154655690785,
+    "producer_headroom": 17.69346490223656,
     "sustainable": true,
-    "buffer_seconds": 60.12615713818413,
+    "buffer_seconds": 60.08244092534108,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 71,
+    "upload_pages_per_commit": 70,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -15,3 +15,4 @@
 {"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}
+{"step": 156800, "loss": 0.6177155636250973, "lr": 0.0003, "grad_norm": 0.1492917735248138, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22948289405, "tokens_per_second": 151428.1687673985, "data_wait_s": 1.3968000075692544e-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 sha256:636047af403283a6ddc680818f6f5718bdffc3d6c4852f90c8c78500c43a6cd6
-size 36171
+oid sha256:2f14411d0de5d7b4540d3074b7620805d94e7006486a20f0a4242c51531c458d
+size 38289