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": 77187.73854260416,
-      "required_pages_per_second_upper_estimate": 0.0053400268526059705
+      "required_token_positions_per_second": 76427.3522811439,
+      "required_pages_per_second_upper_estimate": 0.005287421566698945
     },
     {
       "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": 9648.46731782552,
-      "required_pages_per_second_upper_estimate": 0.004039611783064564
+      "required_token_positions_per_second": 9553.419035142988,
+      "required_pages_per_second_upper_estimate": 0.0039998170519392326
     },
     {
       "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": 19296.93463565104,
-      "required_pages_per_second_upper_estimate": 0.004019042590218104
+      "required_token_positions_per_second": 19106.838070285976,
+      "required_pages_per_second_upper_estimate": 0.003979450488836112
     },
     {
       "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": 28945.40195347656,
-      "required_pages_per_second_upper_estimate": 0.0017352326500412302
+      "required_token_positions_per_second": 28660.257105428966,
+      "required_pages_per_second_upper_estimate": 0.0017181386517917498
     },
     {
       "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": 28945.40195347656,
-      "required_pages_per_second_upper_estimate": 0.0022503776845149235
+      "required_token_positions_per_second": 28660.257105428966,
+      "required_pages_per_second_upper_estimate": 0.0022282089267988596
     },
     {
       "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": 28945.40195347656,
-      "required_pages_per_second_upper_estimate": 0.3712805371079971
+      "required_token_positions_per_second": 28660.257105428966,
+      "required_pages_per_second_upper_estimate": 0.3676230051619267
     }
   ],
   "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": 0.05234745532812557,
-    "producer_p95_seconds": 0.1183577190000733,
-    "gpu_update_seconds": 0.9551776148915861,
-    "producer_batches_per_second": 19.103125333061104,
-    "gpu_batches_per_second": 1.0469257072293316,
-    "producer_headroom": 18.246877692608344,
+    "producer_batch_seconds": 0.048683253679684846,
+    "producer_p95_seconds": 0.1151926549999871,
+    "gpu_update_seconds": 0.9646808086296889,
+    "producer_batches_per_second": 20.54094425527874,
+    "gpu_batches_per_second": 1.0366123085007584,
+    "producer_headroom": 19.815454714199657,
     "sustainable": true,
-    "buffer_seconds": 60.176189738169924,
+    "buffer_seconds": 60.7748909436704,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 34,
+    "upload_pages_per_commit": 42,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791254470898546162-7724d9f6/train.jsonl

@@ -1,2 +1,3 @@
 {"step": 760100, "loss": 0.5428827404975891, "lr": 3e-05, "grad_norm": 0.23948048064094665, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111256501409, "tokens_per_second": 152889.36247226436, "data_wait_s": 1.5780000012455275e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 760200, "loss": 0.5956947207450867, "lr": 3e-05, "grad_norm": 0.2575396636118972, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111271094700, "tokens_per_second": 151912.86003644372, "data_wait_s": 8.612999977231084e-06, "gpu_peak_gb": 89.37899398803711}
+{"step": 760300, "loss": 0.5635800957679749, "lr": 3e-05, "grad_norm": 0.24654580489168332, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111285757983, "tokens_per_second": 153699.24148221326, "data_wait_s": 8.567999998376763e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791254470.1e0c9aa3-e42e-441e-9e17-2e25895f438d-9bhv5.366.0.1791254470898546162-7724d9f6

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a003f48efff857e81781440675d0ba423148c13f48fdc61f1ec4da03b80d8953
-size 4401
+oid sha256:101336f185df18222933af5f179f60a44bbe1ce342ddcca302762d2562c14c8c
+size 6519