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": 77544.36593810296,
-      "required_pages_per_second_upper_estimate": 0.005364699163316141
+      "required_token_positions_per_second": 77654.29730965152,
+      "required_pages_per_second_upper_estimate": 0.005372304470675795
     },
     {
       "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": 9693.04574226287,
-      "required_pages_per_second_upper_estimate": 0.0040582758384731235
+      "required_token_positions_per_second": 9706.78716370644,
+      "required_pages_per_second_upper_estimate": 0.004064029084678036
     },
     {
       "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": 19386.09148452574,
-      "required_pages_per_second_upper_estimate": 0.004037611610614486
+      "required_token_positions_per_second": 19413.57432741288,
+      "required_pages_per_second_upper_estimate": 0.004043335562016572
     },
     {
       "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": 29079.137226788607,
-      "required_pages_per_second_upper_estimate": 0.0017432498754743487
+      "required_token_positions_per_second": 29120.36149111932,
+      "required_pages_per_second_upper_estimate": 0.001745721207175168
     },
     {
       "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": 29079.137226788607,
-      "required_pages_per_second_upper_estimate": 0.0022607750137756347
+      "required_token_positions_per_second": 29120.36149111932,
+      "required_pages_per_second_upper_estimate": 0.0022639800190009723
     },
     {
       "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": 29079.137226788607,
-      "required_pages_per_second_upper_estimate": 0.3729959496002951
+      "required_token_positions_per_second": 29120.36149111932,
+      "required_pages_per_second_upper_estimate": 0.3735247301999631
     }
   ],
   "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": 0.053493828359410855,
-    "producer_p95_seconds": 0.1385831069997039,
-    "gpu_update_seconds": 0.9507847424898769,
-    "producer_batches_per_second": 18.693745253775166,
-    "gpu_batches_per_second": 1.051762775853176,
-    "producer_headroom": 17.77372776728198,
+    "producer_batch_seconds": 0.04716267903125981,
+    "producer_p95_seconds": 0.12978831700002047,
+    "gpu_update_seconds": 0.9494387632664403,
+    "producer_batches_per_second": 21.203206020955506,
+    "gpu_batches_per_second": 1.0532538155063411,
+    "producer_headroom": 20.131145701819534,
     "sustainable": true,
-    "buffer_seconds": 60.85022351935212,
+    "buffer_seconds": 60.76408084905218,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 79,

metrics/void-byte-v1/1791083219122319327-61ae60ac/train.jsonl

@@ -28,3 +28,4 @@
 {"step": 647800, "loss": 0.5407651662826538, "lr": 0.0001, "grad_norm": 0.17395694584160637, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94821128922, "tokens_per_second": 152884.949190817, "data_wait_s": 1.7751000086718705e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 647900, "loss": 0.5759875178337097, "lr": 0.0001, "grad_norm": 0.1969413946572304, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94835801434, "tokens_per_second": 153950.0408605591, "data_wait_s": 8.125000022118911e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 648000, "loss": 0.5908306241035461, "lr": 0.0001, "grad_norm": 0.16779232858622106, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94850495306, "tokens_per_second": 154319.66274522644, "data_wait_s": 1.3393000244832365e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 648100, "loss": 0.5829259157180786, "lr": 0.0001, "grad_norm": 0.21165426190517386, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94865161079, "tokens_per_second": 153921.6830137373, "data_wait_s": 1.3202999980421737e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791083219.1a73f132-ede3-4b51-9e4f-0cafff989449-8p4r7.435.0.1791083219122319327-61ae60ac

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e6a3febd8001147949b01b59f612cc7d4dea02ae257d86a201f09b802812d510
-size 63705
+oid sha256:efceeebf14dedc645f532332292573d861dc9014c678d64c45e808cea0825e40
+size 65823