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": 75759.28414930297,
-      "required_pages_per_second_upper_estimate": 0.006566622739911347
+      "required_token_positions_per_second": 75733.6343305243,
+      "required_pages_per_second_upper_estimate": 0.006564399478628484
     },
     {
       "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": 9469.910518662871,
-      "required_pages_per_second_upper_estimate": 0.003777688893674354
+      "required_token_positions_per_second": 9466.704291315538,
+      "required_pages_per_second_upper_estimate": 0.003776409881648132
     },
     {
       "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": 18939.821037325743,
-      "required_pages_per_second_upper_estimate": 0.004381394267519668
+      "required_token_positions_per_second": 18933.408582631077,
+      "required_pages_per_second_upper_estimate": 0.004379910858981406
     },
     {
       "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": 28409.731555988616,
-      "required_pages_per_second_upper_estimate": 0.0017300035974177588
+      "required_token_positions_per_second": 28400.112873946615,
+      "required_pages_per_second_upper_estimate": 0.0017294178701467255
     },
     {
       "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": 28409.731555988616,
-      "required_pages_per_second_upper_estimate": 0.0021870597783503376
+      "required_token_positions_per_second": 28400.112873946615,
+      "required_pages_per_second_upper_estimate": 0.0021863193055805283
     },
     {
       "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": 28409.731555988616,
-      "required_pages_per_second_upper_estimate": 0.3527537846702586
+      "required_token_positions_per_second": 28400.112873946615,
+      "required_pages_per_second_upper_estimate": 0.3526343528309549
     }
   ],
   "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": 0.03972685024218947,
-    "producer_p95_seconds": 0.13354076500036172,
-    "gpu_update_seconds": 0.973187653868273,
-    "producer_batches_per_second": 25.171892407871066,
-    "gpu_batches_per_second": 1.0275510545424122,
-    "producer_headroom": 24.496974915840635,
+    "producer_batch_seconds": 0.04208974321882408,
+    "producer_p95_seconds": 0.1456406990000687,
+    "gpu_update_seconds": 0.9735172575797549,
+    "producer_batches_per_second": 23.758757443613085,
+    "gpu_batches_per_second": 1.0272031566097588,
+    "producer_headroom": 23.1295603900088,
     "sustainable": true,
-    "buffer_seconds": 60.337634539832926,
+    "buffer_seconds": 60.35806996994481,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 82,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790485060735632541-7579de36/train.jsonl

@@ -34,3 +34,4 @@
 {"step": 88400, "loss": 0.6176230721175671, "lr": 0.0003, "grad_norm": 0.12480244575333811, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12937137368, "tokens_per_second": 150555.39858962144, "data_wait_s": 1.164599962066859e-05, "gpu_peak_gb": 85.46991729736328}
 {"step": 88500, "loss": 0.6561363190412521, "lr": 0.0003, "grad_norm": 0.1666916385605709, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12951962743, "tokens_per_second": 151715.80179587103, "data_wait_s": 7.917999027995393e-06, "gpu_peak_gb": 85.46991729736328}
 {"step": 88600, "loss": 0.6344563355669379, "lr": 0.0003, "grad_norm": 0.13862075684825392, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12966471919, "tokens_per_second": 148368.41707397727, "data_wait_s": 9.277999197365716e-06, "gpu_peak_gb": 85.46991729736328}
+{"step": 88700, "loss": 0.6234282646328211, "lr": 0.0003, "grad_norm": 0.13362376395729664, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 12980961945, "tokens_per_second": 148160.8179684559, "data_wait_s": 1.4209999790182337e-05, "gpu_peak_gb": 85.46991729736328}

runs/void-byte-v1/events.out.tfevents.1790485060.0b09bbb2-f535-42d2-bd00-cbae7fb94f51-r2qrq.28895.0.1790485060735632541-7579de36

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eeb2bd5571a02f7a88735d87c81a958a540cc7b2d8050ef96a163d17ad8c2667
-size 76413
+oid sha256:9ed72efc68b72f19350c57673c81b61f7219bacad35b14558a12141e7e0b8b1d
+size 78531