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": 76163.86771136794,
-      "required_pages_per_second_upper_estimate": 0.0066016910176634755
+      "required_token_positions_per_second": 76060.09615399428,
+      "required_pages_per_second_upper_estimate": 0.006592696362076934
     },
     {
       "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": 9520.483463920993,
-      "required_pages_per_second_upper_estimate": 0.003797863197670733
+      "required_token_positions_per_second": 9507.512019249285,
+      "required_pages_per_second_upper_estimate": 0.003792688694450808
     },
     {
       "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": 19040.966927841986,
-      "required_pages_per_second_upper_estimate": 0.004404792589183726
+      "required_token_positions_per_second": 19015.02403849857,
+      "required_pages_per_second_upper_estimate": 0.004398791158313386
     },
     {
       "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": 28561.450391762977,
-      "required_pages_per_second_upper_estimate": 0.001739242478509205
+      "required_token_positions_per_second": 28522.536057747853,
+      "required_pages_per_second_upper_estimate": 0.0017368727997354155
     },
     {
       "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": 28561.450391762977,
-      "required_pages_per_second_upper_estimate": 0.0021987395143128656
+      "required_token_positions_per_second": 28522.536057747853,
+      "required_pages_per_second_upper_estimate": 0.00219574378115861
     },
     {
       "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": 28561.450391762977,
-      "required_pages_per_second_upper_estimate": 0.3546376248402967
+      "required_token_positions_per_second": 28522.536057747853,
+      "required_pages_per_second_upper_estimate": 0.3541544390497269
     }
   ],
   "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": 0.048936885289068144,
-    "producer_p95_seconds": 0.11113542100019913,
-    "gpu_update_seconds": 0.9680180670367352,
-    "producer_batches_per_second": 20.434484011253303,
-    "gpu_batches_per_second": 1.033038570303927,
-    "producer_headroom": 19.780949713466494,
+    "producer_batch_seconds": 0.05525225264062428,
+    "producer_p95_seconds": 0.12776564699993287,
+    "gpu_update_seconds": 0.9693387693164018,
+    "producer_batches_per_second": 18.098809590701627,
+    "gpu_batches_per_second": 1.03163107847757,
+    "producer_headroom": 17.543877814742608,
     "sustainable": true,
-    "buffer_seconds": 60.017120156277585,
+    "buffer_seconds": 60.099003697616915,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 53,
+    "upload_pages_per_commit": 51,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790606918076636751-3952f1c5/train.jsonl

@@ -4,3 +4,4 @@
 {"step": 195400, "loss": 0.599296510219574, "lr": 0.0003, "grad_norm": 0.13781984487626003, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28598726348, "tokens_per_second": 148718.05140297735, "data_wait_s": 1.5413000028274837e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 195500, "loss": 0.5708072260022163, "lr": 0.0003, "grad_norm": 0.14248767864119002, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28613312760, "tokens_per_second": 149888.48724774213, "data_wait_s": 1.4487000044027809e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 195600, "loss": 0.5437424816191196, "lr": 0.0003, "grad_norm": 0.1409504058238227, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28627969829, "tokens_per_second": 150581.4157509198, "data_wait_s": 1.649299997552589e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 195700, "loss": 0.6062924303114414, "lr": 0.0003, "grad_norm": 0.15676588520254675, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28642713141, "tokens_per_second": 150968.7482119374, "data_wait_s": 1.7213000091942376e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790606918.0a7f8bdc-23d3-4ecb-bd0c-4a4bf460026e-rl2cx.1192.0.1790606918076636751-3952f1c5

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8957ba73196ae04f519a9b2a357252dcaa2491e21fdbf6b0db004fec6b970000
-size 12873
+oid sha256:12f1b1b371629de973b05576b17aa5353f5ff5b8d38e08efbb1e3392cf26e754
+size 14991