Update TensorBoard training metrics

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 1062967.6168812835,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 76166.02571118445,
-      "required_pages_per_second_upper_estimate": 0.00660187806761817
+      "required_token_positions_per_second": 76035.35028777395,
+      "required_pages_per_second_upper_estimate": 0.006590551453110781
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 1206672.263377472,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9520.753213898057,
-      "required_pages_per_second_upper_estimate": 0.003797970804969705
+      "required_token_positions_per_second": 9504.418785971744,
+      "required_pages_per_second_upper_estimate": 0.003791454757448438
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 410042.7581134458,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19041.506427796114,
-      "required_pages_per_second_upper_estimate": 0.004404917393003239
+      "required_token_positions_per_second": 19008.837571943488,
+      "required_pages_per_second_upper_estimate": 0.00439736002815396
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 2163486.5624068812,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28562.25964169417,
-      "required_pages_per_second_upper_estimate": 0.0017392917575842083
+      "required_token_positions_per_second": 28513.25635791523,
+      "required_pages_per_second_upper_estimate": 0.001736307714702431
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 1546694.2737633225,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28562.25964169417,
-      "required_pages_per_second_upper_estimate": 0.002198801812612716
+      "required_token_positions_per_second": 28513.25635791523,
+      "required_pages_per_second_upper_estimate": 0.0021950294041776417
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 1015177.8395898147,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 22,
-      "required_token_positions_per_second": 28562.25964169417,
-      "required_pages_per_second_upper_estimate": 0.3546476730160569
+      "required_token_positions_per_second": 28513.25635791523,
+      "required_pages_per_second_upper_estimate": 0.3540392162349632
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 62,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 688946,
-    "producer_batch_seconds": 0.05994781818750994,
-    "producer_p95_seconds": 0.15240313300000707,
-    "gpu_update_seconds": 0.9679906403357679,
-    "producer_batches_per_second": 16.6811742317646,
-    "gpu_batches_per_second": 1.0330678400497024,
-    "producer_headroom": 16.147220526158325,
+    "producer_batch_seconds": 0.05978618743748143,
+    "producer_p95_seconds": 0.15451852199976202,
+    "gpu_update_seconds": 0.9696542426773701,
+    "producer_batches_per_second": 16.726271449332717,
+    "gpu_batches_per_second": 1.0312954411861701,
+    "producer_headroom": 16.218700075018834,
     "sustainable": true,
-    "buffer_seconds": 60.01541970081761,
+    "buffer_seconds": 60.11856304599694,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 81,
+    "upload_pages_per_commit": 75,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790519810984249790-cb9c9513/train.jsonl

@@ -27,3 +27,4 @@
 {"step": 117700, "loss": 0.6191273359581828, "lr": 0.0003, "grad_norm": 0.16089004303991808, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17223795313, "tokens_per_second": 151289.57461195142, "data_wait_s": 1.3388000297709368e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 117800, "loss": 0.5868233777582645, "lr": 0.0003, "grad_norm": 0.13449859552857432, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17238468392, "tokens_per_second": 150951.86868817345, "data_wait_s": 1.8103000002156477e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 117900, "loss": 0.6241789255291224, "lr": 0.0003, "grad_norm": 0.1445657070540139, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17253039451, "tokens_per_second": 149567.94261848138, "data_wait_s": 1.4555999769072514e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 118000, "loss": 0.5838016532361507, "lr": 0.0003, "grad_norm": 0.1429246888537334, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17267660900, "tokens_per_second": 149406.0071345838, "data_wait_s": 1.2745999811158981e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790519810.cad0b19a-59f2-43f0-8fb4-2a94097843e2-tjh7c.1336.0.1790519810984249790-cb9c9513

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:78feca4ad1b945bb7c09d3be439f09696ec1f043a5bf40ce27056edbb8594d81
-size 61587
+oid sha256:e084e7b48fdbd3ad78d5dfeb98723a635f3b48d8cb604776738d7993d1ba3169
+size 63705