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": 76222.54989778106,
-      "required_pages_per_second_upper_estimate": 0.006606777440853123
+      "required_token_positions_per_second": 76227.96051193927,
+      "required_pages_per_second_upper_estimate": 0.00660724641917528
     },
     {
       "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": 9527.818737222633,
-      "required_pages_per_second_upper_estimate": 0.0038007893478628663
+      "required_token_positions_per_second": 9528.495063992408,
+      "required_pages_per_second_upper_estimate": 0.0038010591447233157
     },
     {
       "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": 19055.637474445266,
-      "required_pages_per_second_upper_estimate": 0.004408186361947593
+      "required_token_positions_per_second": 19056.990127984816,
+      "required_pages_per_second_upper_estimate": 0.0044084992745380795
     },
     {
       "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": 28583.456211667897,
-      "required_pages_per_second_upper_estimate": 0.0017405825174858004
+      "required_token_positions_per_second": 28585.485191977223,
+      "required_pages_per_second_upper_estimate": 0.0017407060717414013
     },
     {
       "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": 28583.456211667897,
-      "required_pages_per_second_upper_estimate": 0.0022004335832451543
+      "required_token_positions_per_second": 28585.485191977223,
+      "required_pages_per_second_upper_estimate": 0.0022005897797659434
     },
     {
       "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": 28583.456211667897,
-      "required_pages_per_second_upper_estimate": 0.3549108634747743
+      "required_token_positions_per_second": 28585.485191977223,
+      "required_pages_per_second_upper_estimate": 0.3549360566196558
     }
   ],
   "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": 0.048124049031239124,
-    "producer_p95_seconds": 0.12490503399999398,
-    "gpu_update_seconds": 0.9672728096721194,
-    "producer_batches_per_second": 20.779631392837757,
-    "gpu_batches_per_second": 1.033834498396553,
-    "producer_headroom": 20.099572441301152,
+    "producer_batch_seconds": 0.05044499264062452,
+    "producer_p95_seconds": 0.1332906779998666,
+    "gpu_update_seconds": 0.9672041532378699,
+    "producer_batches_per_second": 19.823573117040695,
+    "gpu_batches_per_second": 1.0339078845477874,
+    "producer_headroom": 19.173442250816347,
     "sustainable": true,
-    "buffer_seconds": 60.93818700934352,
+    "buffer_seconds": 60.9338616539858,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 68,
+    "upload_pages_per_commit": 71,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790519810984249790-cb9c9513/train.jsonl

@@ -10,3 +10,4 @@
 {"step": 116000, "loss": 0.6013966966420412, "lr": 0.0003, "grad_norm": 0.12939835376412753, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16975010873, "tokens_per_second": 149040.05846871884, "data_wait_s": 1.3589999980467837e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 116100, "loss": 0.5669870059937239, "lr": 0.0003, "grad_norm": 0.1382701306324106, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16989562756, "tokens_per_second": 149802.0055713855, "data_wait_s": 1.2924999964525341e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 116200, "loss": 0.6335533894598484, "lr": 0.0003, "grad_norm": 0.14737405640119294, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17004101407, "tokens_per_second": 149923.8179464614, "data_wait_s": 1.2765999827024643e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 116300, "loss": 0.6132048312574625, "lr": 0.0003, "grad_norm": 0.1570570949934492, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17018720502, "tokens_per_second": 150144.02422067648, "data_wait_s": 1.3435000028039212e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790519810.cad0b19a-59f2-43f0-8fb4-2a94097843e2-tjh7c.1336.0.1790519810984249790-cb9c9513

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:185a25fb8cbc9c0759115a6be717a93193dfa2a4fb1f07eb7f8c56bac0afc92b
-size 25581
+oid sha256:e05b4f6c11676c2f6ff8a726b054b6c9da74478bad9d8ef442a5e26467fe8f5a
+size 27699