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": 77562.03474672964,
-      "required_pages_per_second_upper_estimate": 0.005365921532494219
+      "required_token_positions_per_second": 77510.47250896618,
+      "required_pages_per_second_upper_estimate": 0.005362354337244868
     },
     {
       "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": 9695.254343341205,
-      "required_pages_per_second_upper_estimate": 0.004059200533623787
+      "required_token_positions_per_second": 9688.809063620773,
+      "required_pages_per_second_upper_estimate": 0.004056502029597587
     },
     {
       "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": 19390.50868668241,
-      "required_pages_per_second_upper_estimate": 0.004038531597334266
+      "required_token_positions_per_second": 19377.618127241545,
+      "required_pages_per_second_upper_estimate": 0.004035846833749647
     },
     {
       "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": 29085.763030023616,
-      "required_pages_per_second_upper_estimate": 0.0017436470822612717
+      "required_token_positions_per_second": 29066.427190862316,
+      "required_pages_per_second_upper_estimate": 0.0017424879282276688
     },
     {
       "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": 29085.763030023616,
-      "required_pages_per_second_upper_estimate": 0.0022612901408333285
+      "required_token_positions_per_second": 29066.427190862316,
+      "required_pages_per_second_upper_estimate": 0.0022597868643879056
     },
     {
       "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": 29085.763030023616,
-      "required_pages_per_second_upper_estimate": 0.3730809382899606
+      "required_token_positions_per_second": 29066.427190862316,
+      "required_pages_per_second_upper_estimate": 0.37283291890640596
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.052063578296863255,
-    "producer_p95_seconds": 0.12321040799997718,
-    "gpu_update_seconds": 0.9505681515544394,
-    "producer_batches_per_second": 19.207285259919377,
-    "gpu_batches_per_second": 1.0520024244076829,
-    "producer_headroom": 18.257833645900394,
+    "producer_batch_seconds": 0.047975803968761,
+    "producer_p95_seconds": 0.13067300200009413,
+    "gpu_update_seconds": 0.951200497345328,
+    "producer_batches_per_second": 20.8438403794367,
+    "gpu_batches_per_second": 1.051303066799129,
+    "producer_headroom": 19.82667133550682,
     "sustainable": true,
-    "buffer_seconds": 60.83636169948412,
+    "buffer_seconds": 60.87683183010099,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 64,
+    "upload_pages_per_commit": 65,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790956870018829698-ff939fa5/train.jsonl

@@ -11,3 +11,4 @@
 {"step": 526100, "loss": 0.6050062775611877, "lr": 0.0003, "grad_norm": 0.15781412739970652, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 77003576077, "tokens_per_second": 151795.13748489312, "data_wait_s": 1.161499994850601e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 526200, "loss": 0.5432044863700867, "lr": 0.0003, "grad_norm": 0.1561291186653021, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 77018180283, "tokens_per_second": 153366.29446798816, "data_wait_s": 1.1892000202351483e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 526300, "loss": 0.6389694213867188, "lr": 0.0003, "grad_norm": 0.16326968418129384, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 77032790042, "tokens_per_second": 153197.9701533869, "data_wait_s": 1.2168000012025004e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 526400, "loss": 0.5397037863731384, "lr": 0.0003, "grad_norm": 0.15235547283058473, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 77047436084, "tokens_per_second": 153645.020126853, "data_wait_s": 1.8833000012818957e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1790956870.abdf5fd7-18c6-4e42-b51e-c994b21d5ba1-dh6p5.3284.0.1790956870018829698-ff939fa5

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c0240451ff37af70557dd427917c0394ef6d5f821eadc0e50cc0274031a3458a
-size 27699
+oid sha256:6d35812dc54b419a2f32540a113e15051305f13004ce165f486743fe3fb1e3c5
+size 29817