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": 78849.39479656596,
-      "required_pages_per_second_upper_estimate": 0.005454984087828752
+      "required_token_positions_per_second": 77173.78678960605,
+      "required_pages_per_second_upper_estimate": 0.005339061637961035
     },
     {
       "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": 9856.174349570745,
-      "required_pages_per_second_upper_estimate": 0.004126574379840243
+      "required_token_positions_per_second": 9646.723348700756,
+      "required_pages_per_second_upper_estimate": 0.004038881619610242
     },
     {
       "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": 19712.34869914149,
-      "required_pages_per_second_upper_estimate": 0.004105562384437605
+      "required_token_positions_per_second": 19293.446697401512,
+      "required_pages_per_second_upper_estimate": 0.004018316144663844
     },
     {
       "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": 29568.523048712235,
-      "required_pages_per_second_upper_estimate": 0.0017725878082497874
+      "required_token_positions_per_second": 28940.17004610227,
+      "required_pages_per_second_upper_estimate": 0.0017349190051828016
     },
     {
       "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": 29568.523048712235,
-      "required_pages_per_second_upper_estimate": 0.002298822608849458
+      "required_token_positions_per_second": 28940.17004610227,
+      "required_pages_per_second_upper_estimate": 0.002249970926729301
     },
     {
       "source": "cortex",
@@ -91,9 +91,9 @@
       "read_seconds": 0.04061180899998362,
       "tokens_per_second": 910315.184299418,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
-      "estimated_pages_for_buffer": 24,
-      "required_token_positions_per_second": 29568.523048712235,
-      "required_pages_per_second_upper_estimate": 0.3792732654623752
+      "estimated_pages_for_buffer": 23,
+      "required_token_positions_per_second": 28940.17004610227,
+      "required_pages_per_second_upper_estimate": 0.37121342781778416
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,27 +155,27 @@
   "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
   "estimated_fresh_generation_batch_seconds": 2.558590814491007,
   "selected": {
-    "prefetch_batches": 65,
+    "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.04021698036923453,
+    "producer_batch_seconds": 0.06926162537500291,
     "producer_p95_seconds": 0.1268318909999948,
-    "gpu_update_seconds": 0.9350483943500222,
-    "producer_batches_per_second": 24.86511893282239,
-    "gpu_batches_per_second": 1.0694633625836312,
-    "producer_headroom": 23.250089533457913,
+    "gpu_update_seconds": 0.9553502953146503,
+    "producer_batches_per_second": 14.438009425648104,
+    "gpu_batches_per_second": 1.046736474468398,
+    "producer_headroom": 13.793356568548623,
     "sustainable": true,
-    "buffer_seconds": 60.77814563275144,
+    "buffer_seconds": 60.18706860482297,
     "buffer_drain_seconds": null,
-    "shard_target_mib": 16,
-    "upload_pages_per_commit": 23,
+    "shard_target_mib": 32,
+    "upload_pages_per_commit": 61,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 412574.99951053865,
+    "generation_bytes_per_second": 1081743.9514613075,
     "upload_bytes_per_second": 85005.02186575368,
     "upload_sustainable": false,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 88.52899064051742,
+    "estimated_full_shards_per_hour": 116.05853513660153,
     "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
   },
   "upload_measurements": [

metrics/void-byte-v1/1791039222312944374-63c10d34/train.jsonl

@@ -1 +1,2 @@
 {"step": 604500, "loss": 0.636208713054657, "lr": 0.0001, "grad_norm": 0.16678688842682265, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 88481063206, "tokens_per_second": 152754.29312491545, "data_wait_s": 1.4469000007011346e-05, "gpu_peak_gb": 89.37861251831055}
+{"step": 604600, "loss": 0.5752429366111755, "lr": 0.0001, "grad_norm": 0.17567589109079268, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 88495546086, "tokens_per_second": 149962.14570994265, "data_wait_s": 1.381399999900168e-05, "gpu_peak_gb": 89.37861251831055}

runs/void-byte-v1/events.out.tfevents.1791039222.f2a393d4-e1e4-407b-a6aa-034aa2ada5ac-vwfgg.353.0.1791039222312944374-63c10d34

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7e77d6fa6e8786044746245471be3e87e95daa5bfbd30666f2bcbeace3ec6981
-size 2283
+oid sha256:8256db8c9b4f8c98e7a0714b0a352d6d3323c7ee1fec0a9f707d5990f51886f2
+size 4401