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": 76659.54058607723,
-      "required_pages_per_second_upper_estimate": 0.005303484892385325
+      "required_token_positions_per_second": 76862.24765575059,
+      "required_pages_per_second_upper_estimate": 0.005317508637810531
     },
     {
       "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": 9582.442573259654,
-      "required_pages_per_second_upper_estimate": 0.0040119686012682855
+      "required_token_positions_per_second": 9607.780956968823,
+      "required_pages_per_second_upper_estimate": 0.004022577253401694
     },
     {
       "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": 19164.885146519307,
-      "required_pages_per_second_upper_estimate": 0.003991540164011173
+      "required_token_positions_per_second": 19215.561913937647,
+      "required_pages_per_second_upper_estimate": 0.004002094798228184
     },
     {
       "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": 28747.327719778965,
-      "required_pages_per_second_upper_estimate": 0.0017233584021729023
+      "required_token_positions_per_second": 28823.34287090647,
+      "required_pages_per_second_upper_estimate": 0.0017279153944145809
     },
     {
       "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": 28747.327719778965,
-      "required_pages_per_second_upper_estimate": 0.0022349782840814145
+      "required_token_positions_per_second": 28823.34287090647,
+      "required_pages_per_second_upper_estimate": 0.0022408881277262635
     },
     {
       "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": 28747.327719778965,
-      "required_pages_per_second_upper_estimate": 0.3687398535136666
+      "required_token_positions_per_second": 28823.34287090647,
+      "required_pages_per_second_upper_estimate": 0.36971489425362
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.0464060150468768,
-    "producer_p95_seconds": 0.14688474200011115,
-    "gpu_update_seconds": 0.9617589596328776,
-    "producer_batches_per_second": 21.548930650258484,
-    "gpu_batches_per_second": 1.0397615639387645,
-    "producer_headroom": 20.72487712339363,
+    "producer_batch_seconds": 0.060600805968739735,
+    "producer_p95_seconds": 0.14303328200003307,
+    "gpu_update_seconds": 0.959222534451657,
+    "producer_batches_per_second": 16.501430699054385,
+    "gpu_batches_per_second": 1.0425109545322073,
+    "producer_headroom": 15.828544177225323,
     "sustainable": true,
-    "buffer_seconds": 60.59081445687129,
+    "buffer_seconds": 60.43101967045439,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 85,
+    "upload_pages_per_commit": 79,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791300814123050567-72be911d/train.jsonl

@@ -29,3 +29,4 @@
 {"step": 802900, "loss": 0.5392683744430542, "lr": 3e-05, "grad_norm": 0.28495113546919615, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117521292308, "tokens_per_second": 150503.1734724028, "data_wait_s": 1.4216999261407182e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 803000, "loss": 0.5339804291725159, "lr": 3e-05, "grad_norm": 0.27437309195666915, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117535845279, "tokens_per_second": 151413.72738506095, "data_wait_s": 1.5567999980703462e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 803100, "loss": 0.5562798976898193, "lr": 3e-05, "grad_norm": 0.27674891289864345, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117550351913, "tokens_per_second": 151145.0404269744, "data_wait_s": 1.3466000382322818e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 803200, "loss": 0.5094635486602783, "lr": 3e-05, "grad_norm": 0.26077655390242566, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117565077044, "tokens_per_second": 153358.40296764378, "data_wait_s": 1.6062000213423744e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791300814.0c9d201b-2dbc-40fb-b739-4ee2350a0f54-7b2tx.840.0.1791300814123050567-72be911d

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ecd594b3f4c6f3f352f4dbd02d2e30ada2a820b8ec1add7b0131decbd15f0350
-size 65823
+oid sha256:3fd338d2488416009ba9f4098ea683f575ef197e0f9ec6b15054a836ba49933e
+size 67941