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": 76532.53358185262,
-      "required_pages_per_second_upper_estimate": 0.005294698253136216
+      "required_token_positions_per_second": 76807.44846618304,
+      "required_pages_per_second_upper_estimate": 0.00531371750272461
     },
     {
       "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": 9566.566697731578,
-      "required_pages_per_second_upper_estimate": 0.00400532170371066
+      "required_token_positions_per_second": 9600.93105827288,
+      "required_pages_per_second_upper_estimate": 0.004019709343859853
     },
     {
       "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": 19133.133395463155,
-      "required_pages_per_second_upper_estimate": 0.003984927111616161
+      "required_token_positions_per_second": 19201.86211654576,
+      "required_pages_per_second_upper_estimate": 0.003999241491719407
     },
     {
       "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": 28699.700093194733,
-      "required_pages_per_second_upper_estimate": 0.0017205031986823524
+      "required_token_positions_per_second": 28802.793174818642,
+      "required_pages_per_second_upper_estimate": 0.0017266834715116849
     },
     {
       "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": 28699.700093194733,
-      "required_pages_per_second_upper_estimate": 0.0022312754456062798
+      "required_token_positions_per_second": 28802.793174818642,
+      "required_pages_per_second_upper_estimate": 0.0022392904792439957
     },
     {
       "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": 28699.700093194733,
-      "required_pages_per_second_upper_estimate": 0.36812893745840525
+      "required_token_positions_per_second": 28802.793174818642,
+      "required_pages_per_second_upper_estimate": 0.3694513048167499
     }
   ],
   "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": 0.050549856125009285,
-    "producer_p95_seconds": 0.1263461799999277,
-    "gpu_update_seconds": 0.9633550145200259,
-    "producer_batches_per_second": 19.782449974279057,
-    "gpu_batches_per_second": 1.0380389211948327,
-    "producer_headroom": 19.05752238221329,
+    "producer_batch_seconds": 0.06011445402344684,
+    "producer_p95_seconds": 0.1275092969999605,
+    "gpu_update_seconds": 0.959906903201727,
+    "producer_batches_per_second": 16.63493441377615,
+    "gpu_batches_per_second": 1.0417676929549566,
+    "producer_headroom": 15.9679883780917,
     "sustainable": true,
-    "buffer_seconds": 60.69136591476163,
+    "buffer_seconds": 60.4741349017088,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 71,
+    "upload_pages_per_commit": 68,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -8,3 +8,4 @@
 {"step": 720800, "loss": 0.495766818523407, "lr": 5e-05, "grad_norm": 0.21895161647899114, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105504891856, "tokens_per_second": 151486.92475880412, "data_wait_s": 1.476499983255053e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 720900, "loss": 0.5261343717575073, "lr": 5e-05, "grad_norm": 0.21882450318529473, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105519523252, "tokens_per_second": 151969.8927038483, "data_wait_s": 1.5592000181641197e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 721000, "loss": 0.5623640418052673, "lr": 5e-05, "grad_norm": 0.21036169801123655, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105534247906, "tokens_per_second": 153168.9143081392, "data_wait_s": 1.0492999990674434e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 721100, "loss": 0.5602995157241821, "lr": 5e-05, "grad_norm": 0.21492721161542766, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105548900762, "tokens_per_second": 152262.951680818, "data_wait_s": 1.646800001253723e-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:dee83512be7f474234aef88d0fa446cd12ec8cae5eb32e71bfeafd0861124f47
-size 21345
+oid sha256:a70adce92dd1b67ea8c65353a6fdc2f8e5fd737d671ad8ac22299a10486cbe1c
+size 23463