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": 77220.10362638804,
-      "required_pages_per_second_upper_estimate": 0.0053422659441994804
+      "required_token_positions_per_second": 77023.52964646976,
+      "required_pages_per_second_upper_estimate": 0.00532866650015426
     },
     {
       "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": 9652.512953298505,
-      "required_pages_per_second_upper_estimate": 0.004041305606154627
+      "required_token_positions_per_second": 9627.94120580872,
+      "required_pages_per_second_upper_estimate": 0.00403101792859709
     },
     {
       "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": 19305.02590659701,
-      "required_pages_per_second_upper_estimate": 0.004020727788574986
+      "required_token_positions_per_second": 19255.88241161744,
+      "required_pages_per_second_upper_estimate": 0.004010492494571856
     },
     {
       "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": 28957.538859895514,
-      "required_pages_per_second_upper_estimate": 0.0017359602390490635
+      "required_token_positions_per_second": 28883.82361742616,
+      "required_pages_per_second_upper_estimate": 0.0017315411228196835
     },
     {
       "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": 28957.538859895514,
-      "required_pages_per_second_upper_estimate": 0.0022513212756043875
+      "required_token_positions_per_second": 28883.82361742616,
+      "required_pages_per_second_upper_estimate": 0.00224559023974148
     },
     {
       "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": 28957.538859895514,
-      "required_pages_per_second_upper_estimate": 0.3714362163119446
+      "required_token_positions_per_second": 28883.82361742616,
+      "required_pages_per_second_upper_estimate": 0.37049067633080845
     }
   ],
   "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": 0.043973964898423645,
-    "producer_p95_seconds": 0.12237705600000481,
-    "gpu_update_seconds": 0.9547772735027165,
-    "producer_batches_per_second": 22.740728572234055,
-    "gpu_batches_per_second": 1.0473646867728412,
-    "producer_headroom": 21.712330823662953,
+    "producer_batch_seconds": 0.06252793966406145,
+    "producer_p95_seconds": 0.13424668199968437,
+    "gpu_update_seconds": 0.9572139882241711,
+    "producer_batches_per_second": 15.992850642010836,
+    "gpu_batches_per_second": 1.0446984815330642,
+    "producer_headroom": 15.308580346112688,
     "sustainable": true,
-    "buffer_seconds": 60.15096823067114,
+    "buffer_seconds": 60.304481258122784,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 88,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791463982175382982-d4a3b372/train.jsonl

@@ -23,3 +23,4 @@
 {"step": 927300, "loss": 0.5016880631446838, "lr": 3e-05, "grad_norm": 0.3294710496400265, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135730323364, "tokens_per_second": 152485.7517690447, "data_wait_s": 8.615999831818044e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 927400, "loss": 0.500203013420105, "lr": 3e-05, "grad_norm": 0.32786574559242104, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135745010409, "tokens_per_second": 154030.78989010258, "data_wait_s": 1.0583999937807675e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 927500, "loss": 0.5234514474868774, "lr": 3e-05, "grad_norm": 0.3416548070371066, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135759655387, "tokens_per_second": 153521.68107448114, "data_wait_s": 7.235999873955734e-06, "gpu_peak_gb": 89.37899398803711}
+{"step": 927600, "loss": 0.5498512983322144, "lr": 3e-05, "grad_norm": 0.3315312814857236, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135774307827, "tokens_per_second": 153870.60190463637, "data_wait_s": 1.0039000244432827e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791463982.2e076f0c-3fe2-4316-8f17-4e30b2e706e2-b5972.360.0.1791463982175382982-d4a3b372

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d49bdf3e2ef1702b6ae9e5966b928df0e51af3ecf97b0af41adce0fcf2b958bb
-size 53115
+oid sha256:92538b8649dab079389323fa93a6686f0529d90eda07f92025aff286cd26af18
+size 55233