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": 76624.51111715428,
-      "required_pages_per_second_upper_estimate": 0.0053010614724456695
+      "required_token_positions_per_second": 76616.11044688022,
+      "required_pages_per_second_upper_estimate": 0.00530048029458387
     },
     {
       "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": 9578.063889644285,
-      "required_pages_per_second_upper_estimate": 0.004010135337875842
+      "required_token_positions_per_second": 9577.013805860028,
+      "required_pages_per_second_upper_estimate": 0.0040096956897236165
     },
     {
       "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": 19156.12777928857,
-      "required_pages_per_second_upper_estimate": 0.003989716235364315
+      "required_token_positions_per_second": 19154.027611720056,
+      "required_pages_per_second_upper_estimate": 0.0039892788258449365
     },
     {
       "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": 28734.191668932854,
-      "required_pages_per_second_upper_estimate": 0.001722570915982266
+      "required_token_positions_per_second": 28731.041417580083,
+      "required_pages_per_second_upper_estimate": 0.0017223820632238228
     },
     {
       "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": 28734.191668932854,
-      "required_pages_per_second_upper_estimate": 0.0022339570139075105
+      "required_token_positions_per_second": 28731.041417580083,
+      "required_pages_per_second_upper_estimate": 0.0022337120957213238
     },
     {
       "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": 28734.191668932854,
-      "required_pages_per_second_upper_estimate": 0.36857135835780525
+      "required_token_positions_per_second": 28731.041417580083,
+      "required_pages_per_second_upper_estimate": 0.36853095031592825
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,24 +158,24 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.10991098371079033,
+    "producer_batch_seconds": 0.08602275621095146,
     "producer_p95_seconds": 0.24481981599819846,
-    "gpu_update_seconds": 0.9621986349416874,
-    "producer_batches_per_second": 9.098271767190331,
-    "gpu_batches_per_second": 1.039286446358972,
-    "producer_headroom": 8.75434467471903,
+    "gpu_update_seconds": 0.9623041364272515,
+    "producer_batches_per_second": 11.624830963887334,
+    "gpu_batches_per_second": 1.0391725049761316,
+    "producer_headroom": 11.186622921816376,
     "sustainable": true,
-    "buffer_seconds": 60.61851400132631,
+    "buffer_seconds": 60.62516059491685,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 86,
+    "upload_pages_per_commit": 85,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 341771.32780130504,
+    "generation_bytes_per_second": 341829.7152220353,
     "upload_bytes_per_second": 5551466.913995439,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 73.33617091683735,
+    "estimated_full_shards_per_hour": 73.34869949813647,
     "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-v1/1790914434605211575-090606d4/train.jsonl

@@ -189,3 +189,4 @@
 {"step": 503900, "loss": 0.6091482639312744, "lr": 0.0003, "grad_norm": 0.15232177667237645, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 73754158701, "tokens_per_second": 153796.19281024934, "data_wait_s": 9.94700167211704e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 504000, "loss": 0.5923633575439453, "lr": 0.0003, "grad_norm": 0.16738223139786748, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 73768891949, "tokens_per_second": 153498.19427815807, "data_wait_s": 1.2784999853465706e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 504100, "loss": 0.5700789093971252, "lr": 0.0003, "grad_norm": 0.15497567197623596, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 73783650213, "tokens_per_second": 153380.6560250495, "data_wait_s": 1.034199885907583e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 504200, "loss": 0.5707831382751465, "lr": 0.0003, "grad_norm": 0.16435508490881467, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 73798294014, "tokens_per_second": 152955.56946700104, "data_wait_s": 1.286200131289661e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1790914434.4dec8a9d-3312-413a-9119-9058bec1adb7-bx5st.8214.0.1790914434605211575-090606d4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:698ce0ecd02b2872d5238302aee6547796b03649a247d475147c2c66e8953716
-size 407052
+oid sha256:659ab78ba8e50343ecabe4909e8acba64b341fd371efaf104fff7f734e4d5f97
+size 409170