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": 76361.54687216604,
-      "required_pages_per_second_upper_estimate": 0.006618825346308203
+      "required_token_positions_per_second": 76302.14292823436,
+      "required_pages_per_second_upper_estimate": 0.006613676363006123
     },
     {
       "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": 9545.193359020755,
-      "required_pages_per_second_upper_estimate": 0.003807720344271883
+      "required_token_positions_per_second": 9537.767866029295,
+      "required_pages_per_second_upper_estimate": 0.0038047582040965752
     },
     {
       "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": 19090.38671804151,
-      "required_pages_per_second_upper_estimate": 0.004416224987887785
+      "required_token_positions_per_second": 19075.53573205859,
+      "required_pages_per_second_upper_estimate": 0.004412789473649063
     },
     {
       "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": 28635.580077062263,
-      "required_pages_per_second_upper_estimate": 0.0017437565874155307
+      "required_token_positions_per_second": 28613.303598087885,
+      "required_pages_per_second_upper_estimate": 0.0017424000667216435
     },
     {
       "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": 28635.580077062263,
-      "required_pages_per_second_upper_estimate": 0.002204446222691307
+      "required_token_positions_per_second": 28613.303598087885,
+      "required_pages_per_second_upper_estimate": 0.0022027313176746187
     },
     {
       "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": 28635.580077062263,
-      "required_pages_per_second_upper_estimate": 0.3555580674356167
+      "required_token_positions_per_second": 28613.303598087885,
+      "required_pages_per_second_upper_estimate": 0.3552814681213341
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,14 +158,14 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 690670,
-    "producer_batch_seconds": 0.05416851377343335,
-    "producer_p95_seconds": 0.16440354000042134,
-    "gpu_update_seconds": 0.9655121330036077,
-    "producer_batches_per_second": 18.460908936557246,
-    "gpu_batches_per_second": 1.0357197655187451,
-    "producer_headroom": 17.82423156452075,
+    "producer_batch_seconds": 0.056468398945305864,
+    "producer_p95_seconds": 0.1590126129995042,
+    "gpu_update_seconds": 0.9662638186891361,
+    "producer_batches_per_second": 17.709019888603173,
+    "gpu_batches_per_second": 1.0349140479632482,
+    "producer_headroom": 17.111585182803562,
     "sustainable": true,
-    "buffer_seconds": 60.82726437922729,
+    "buffer_seconds": 60.87462057741558,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
     "upload_pages_per_commit": 78,

metrics/void-byte-v1/1790696019647648547-3fdea0a8/train.jsonl

@@ -43,3 +43,4 @@
 {"step": 279300, "loss": 0.5953002441674471, "lr": 0.0003, "grad_norm": 0.1420977077373414, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40882067628, "tokens_per_second": 150461.6564545454, "data_wait_s": 1.4311000086308923e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 279400, "loss": 0.5541030168533325, "lr": 0.0003, "grad_norm": 0.1464481355640544, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40896558348, "tokens_per_second": 149532.5978732615, "data_wait_s": 1.0278999980073422e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 279500, "loss": 0.597193019464612, "lr": 0.0003, "grad_norm": 0.1399536693597878, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40911096080, "tokens_per_second": 149875.69336045466, "data_wait_s": 1.2569000318762846e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 279600, "loss": 0.6400405652821064, "lr": 0.0003, "grad_norm": 0.15598653516577976, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40925686562, "tokens_per_second": 150712.69561718983, "data_wait_s": 1.082600010704482e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790696019.3e6d470a-8cc7-4d7e-a224-51a0f9a2ce55-85btc.22407.0.1790696019647648547-3fdea0a8

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:acacf3adcde0526c712108d1f072a0eeeed39c5cb09dd90a242e1663c2416622
-size 95475
+oid sha256:965c277a6b42276ffcfe3b3dfd907d6d15eef7041ab5e611264916a77c29154e
+size 97593