Update TensorBoard training metrics

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 598315.063954628,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 77431.05705095429,
-      "required_pages_per_second_upper_estimate": 0.005356860191590355
+      "required_token_positions_per_second": 77454.1153760207,
+      "required_pages_per_second_upper_estimate": 0.005358455420021135
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 8533.957213384612,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9678.882131369286,
-      "required_pages_per_second_upper_estimate": 0.0040523458303618086
+      "required_token_positions_per_second": 9681.764422002587,
+      "required_pages_per_second_upper_estimate": 0.004053552585260899
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 380144.35761751566,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19357.764262738572,
-      "required_pages_per_second_upper_estimate": 0.004031711797355294
+      "required_token_positions_per_second": 19363.528844005174,
+      "required_pages_per_second_upper_estimate": 0.004032912407610896
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 43742.16556775435,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29036.64639410786,
-      "required_pages_per_second_upper_estimate": 0.0017407026149348827
+      "required_token_positions_per_second": 29045.29326600776,
+      "required_pages_per_second_upper_estimate": 0.0017412209817022728
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 22018.86276523721,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 29036.64639410786,
-      "required_pages_per_second_upper_estimate": 0.0022574715384321275
+      "required_token_positions_per_second": 29045.29326600776,
+      "required_pages_per_second_upper_estimate": 0.002258143794688775
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 910315.184299418,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 23,
-      "required_token_positions_per_second": 29036.64639410786,
-      "required_pages_per_second_upper_estimate": 0.37245092282176806
+      "required_token_positions_per_second": 29045.29326600776,
+      "required_pages_per_second_upper_estimate": 0.3725618356102123
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,14 +158,14 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.06774286313282474,
-    "producer_p95_seconds": 0.22404742099979558,
-    "gpu_update_seconds": 0.9521760751823721,
-    "producer_batches_per_second": 14.761702617134453,
-    "gpu_batches_per_second": 1.0502259257128133,
-    "producer_headroom": 14.055740060992434,
+    "producer_batch_seconds": 0.05481918577343947,
+    "producer_p95_seconds": 0.19849216200009323,
+    "gpu_update_seconds": 0.9518926094768325,
+    "producer_batches_per_second": 18.241788634600837,
+    "gpu_batches_per_second": 1.0505386742624334,
+    "producer_headroom": 17.364223784915016,
     "sustainable": true,
-    "buffer_seconds": 60.939268811671816,
+    "buffer_seconds": 60.92112700651728,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 83,

metrics/void-byte-v1/1791463982175382982-d4a3b372/train.jsonl

@@ -30,3 +30,4 @@
 {"step": 928000, "loss": 0.5244894027709961, "lr": 3e-05, "grad_norm": 0.34271158779188887, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135833040179, "tokens_per_second": 153403.62911481096, "data_wait_s": 1.130600003307336e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 928100, "loss": 0.5575608015060425, "lr": 3e-05, "grad_norm": 0.35268365170265237, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135847590762, "tokens_per_second": 152781.83261242192, "data_wait_s": 6.800999926781515e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 928200, "loss": 0.5369636416435242, "lr": 3e-05, "grad_norm": 0.32165364746021524, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135862068058, "tokens_per_second": 152026.8280368941, "data_wait_s": 6.794999990233919e-06, "gpu_peak_gb": 89.37899398803711}
+{"step": 928300, "loss": 0.5516103506088257, "lr": 3e-05, "grad_norm": 0.328012424760905, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135876730454, "tokens_per_second": 153856.77058871454, "data_wait_s": 1.1735000043699984e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791463982.2e076f0c-3fe2-4316-8f17-4e30b2e706e2-b5972.360.0.1791463982175382982-d4a3b372

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ad2abfeaf2733512ac7c1312b6e5dc47aa3a96786d6c30e4eef2d987ea3d1b47
-size 67941
+oid sha256:18873e73d6fad262097f185301734dc015284d16643899fda26c2efd3c347023
+size 70059