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": 76798.62924967271,
-      "required_pages_per_second_upper_estimate": 0.0053131073688630235
+      "required_token_positions_per_second": 76829.40532140016,
+      "required_pages_per_second_upper_estimate": 0.005315236528915454
     },
     {
       "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": 9599.828656209089,
-      "required_pages_per_second_upper_estimate": 0.004019247791136517
+      "required_token_positions_per_second": 9603.67566517502,
+      "required_pages_per_second_upper_estimate": 0.00402085845345587
     },
     {
       "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": 19199.657312418178,
-      "required_pages_per_second_upper_estimate": 0.003998782289164227
+      "required_token_positions_per_second": 19207.35133035004,
+      "required_pages_per_second_upper_estimate": 0.004000384750194536
     },
     {
       "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": 28799.48596862727,
-      "required_pages_per_second_upper_estimate": 0.0017264852095503233
+      "required_token_positions_per_second": 28811.026995525062,
+      "required_pages_per_second_upper_estimate": 0.0017271770764907184
     },
     {
       "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": 28799.48596862727,
-      "required_pages_per_second_upper_estimate": 0.002239033358509479
+      "required_token_positions_per_second": 28811.026995525062,
+      "required_pages_per_second_upper_estimate": 0.00223993062258717
     },
     {
       "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": 28799.48596862727,
-      "required_pages_per_second_upper_estimate": 0.36940888352672835
+      "required_token_positions_per_second": 28811.026995525062,
+      "required_pages_per_second_upper_estimate": 0.369556919427984
     }
   ],
   "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": 720750,
-    "producer_batch_seconds": 0.049861851234400234,
-    "producer_p95_seconds": 0.12851170600015394,
-    "gpu_update_seconds": 0.9600171346849162,
-    "producer_batches_per_second": 20.05541260991307,
-    "gpu_batches_per_second": 1.0416480746754655,
-    "producer_headroom": 19.253539748692482,
+    "producer_batch_seconds": 0.05445190348437379,
+    "producer_p95_seconds": 0.12841596800035404,
+    "gpu_update_seconds": 0.9596325741631597,
+    "producer_batches_per_second": 18.364830905993447,
+    "gpu_batches_per_second": 1.042065501863609,
+    "producer_headroom": 17.623489956389644,
     "sustainable": true,
-    "buffer_seconds": 60.48107948514972,
+    "buffer_seconds": 60.45685217227906,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 77,
+    "upload_pages_per_commit": 85,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791209746770022608-f3fe5fb7/train.jsonl

@@ -17,3 +17,4 @@
 {"step": 721700, "loss": 0.5054534673690796, "lr": 5e-05, "grad_norm": 0.21090170174195838, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105636903287, "tokens_per_second": 152603.45544524852, "data_wait_s": 1.4651000128651503e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 721800, "loss": 0.597263753414154, "lr": 5e-05, "grad_norm": 0.21663511348799927, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105651647851, "tokens_per_second": 153471.2782438447, "data_wait_s": 1.55860002450936e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 721900, "loss": 0.5127010941505432, "lr": 5e-05, "grad_norm": 0.2314292797174232, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105666361451, "tokens_per_second": 153174.7397481044, "data_wait_s": 7.755999831715599e-06, "gpu_peak_gb": 89.37899398803711}
+{"step": 722000, "loss": 0.5648269057273865, "lr": 5e-05, "grad_norm": 0.2197077038734712, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105681078705, "tokens_per_second": 152364.29076591463, "data_wait_s": 1.9053999949392164e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791209746.ebd20fd8-0561-42d8-bf2b-56d08e7106ef-6qqk2.329.0.1791209746770022608-f3fe5fb7

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:703721de13980897b1e2939552b2835dffa7e08f3261b7c3e67610d11be982f2
-size 40407
+oid sha256:f5588aa3e84735b250a9a7786e6cfd19c01f13c01352c9a6e306466cb6453238
+size 42525