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": 75202.24499235753,
-      "required_pages_per_second_upper_estimate": 0.006518339997590153
+      "required_token_positions_per_second": 75794.85813714447,
+      "required_pages_per_second_upper_estimate": 0.0065697062030160005
     },
     {
       "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": 9400.280624044692,
-      "required_pages_per_second_upper_estimate": 0.0037499124876514646
+      "required_token_positions_per_second": 9474.35726714306,
+      "required_pages_per_second_upper_estimate": 0.003779462768127916
     },
     {
       "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": 18800.561248089383,
-      "required_pages_per_second_upper_estimate": 0.004349178966168419
+      "required_token_positions_per_second": 18948.71453428612,
+      "required_pages_per_second_upper_estimate": 0.004383451621521251
     },
     {
       "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": 28200.84187213407,
-      "required_pages_per_second_upper_estimate": 0.0017172833116304881
+      "required_token_positions_per_second": 28423.07180142918,
+      "required_pages_per_second_upper_estimate": 0.0017308159483742308
     },
     {
       "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": 28200.84187213407,
-      "required_pages_per_second_upper_estimate": 0.0021709788722435604
+      "required_token_positions_per_second": 28423.07180142918,
+      "required_pages_per_second_upper_estimate": 0.002188086747372515
     },
     {
       "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": 28200.84187213407,
-      "required_pages_per_second_upper_estimate": 0.3501600739055847
+      "required_token_positions_per_second": 28423.07180142918,
+      "required_pages_per_second_upper_estimate": 0.352919425871701
     }
   ],
   "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": 0.05189300645313466,
-    "producer_p95_seconds": 0.15040141300005416,
-    "gpu_update_seconds": 0.9803962635356518,
-    "producer_batches_per_second": 19.270419433168026,
-    "gpu_batches_per_second": 1.0199957274354048,
-    "producer_headroom": 18.892647209042746,
+    "producer_batch_seconds": 0.055870519976537025,
+    "producer_p95_seconds": 0.1529188360000262,
+    "gpu_update_seconds": 0.9727308924649655,
+    "producer_batches_per_second": 17.898526815572016,
+    "gpu_batches_per_second": 1.0280335576327104,
+    "producer_headroom": 17.410449963119483,
     "sustainable": true,
-    "buffer_seconds": 60.78456833921041,
+    "buffer_seconds": 60.30931533282786,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 82,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -29,3 +29,4 @@
 {"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}
 {"step": 118100, "loss": 0.6587215159088373, "lr": 0.0003, "grad_norm": 0.16140686632219703, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17282311564, "tokens_per_second": 150762.70017903758, "data_wait_s": 1.4431000181502895e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 118200, "loss": 0.5931202527135611, "lr": 0.0003, "grad_norm": 0.1292053724103761, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17296945848, "tokens_per_second": 150656.84809803465, "data_wait_s": 1.3708000096812611e-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 sha256:5e7987f9d6819207c07891473bd320c3e9b540357e0741f7bbebc46feea6b7b1
-size 65823
+oid sha256:0765dc4caa7fb0c17fea7b671efa1ff65054b76b3aef29e9e65c306f821483d8
+size 67941