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": 77066.59893697694,
-      "required_pages_per_second_upper_estimate": 0.005331646133606069
+      "required_token_positions_per_second": 76685.95931588428,
+      "required_pages_per_second_upper_estimate": 0.0053053126014132755
     },
     {
       "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": 9633.324867122117,
-      "required_pages_per_second_upper_estimate": 0.004033271955165377
+      "required_token_positions_per_second": 9585.744914485535,
+      "required_pages_per_second_upper_estimate": 0.004013351222578835
     },
     {
       "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": 19266.649734244234,
-      "required_pages_per_second_upper_estimate": 0.004012735043921625
+      "required_token_positions_per_second": 19171.48982897107,
+      "required_pages_per_second_upper_estimate": 0.003992915745188686
     },
     {
       "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": 28899.97460136635,
-      "required_pages_per_second_upper_estimate": 0.0017325093496457734
+      "required_token_positions_per_second": 28757.234743456604,
+      "required_pages_per_second_upper_estimate": 0.0017239523131152275
     },
     {
       "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": 28899.97460136635,
-      "required_pages_per_second_upper_estimate": 0.0022468459077021594
+      "required_token_positions_per_second": 28757.234743456604,
+      "required_pages_per_second_upper_estimate": 0.002235748511595959
     },
     {
       "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": 28899.97460136635,
-      "required_pages_per_second_upper_estimate": 0.3706978438112178
+      "required_token_positions_per_second": 28757.234743456604,
+      "required_pages_per_second_upper_estimate": 0.36886693017607014
     }
   ],
   "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": 0.057243127031203045,
-    "producer_p95_seconds": 0.137696139000127,
-    "gpu_update_seconds": 0.956679041465069,
-    "producer_batches_per_second": 17.469346135736142,
-    "gpu_batches_per_second": 1.0452826461721048,
-    "producer_headroom": 16.71255731615756,
+    "producer_batch_seconds": 0.05727790491404505,
+    "producer_p95_seconds": 0.1350072000000182,
+    "gpu_update_seconds": 0.9614276284436911,
+    "producer_batches_per_second": 17.45873913336504,
+    "gpu_batches_per_second": 1.0401198908946978,
+    "producer_headroom": 16.785314160608213,
     "sustainable": true,
-    "buffer_seconds": 60.27077961229935,
+    "buffer_seconds": 60.56994059195254,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 84,
+    "upload_pages_per_commit": 78,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -21,3 +21,4 @@
 {"step": 802100, "loss": 0.48784488439559937, "lr": 3e-05, "grad_norm": 0.2600461222229231, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117404298080, "tokens_per_second": 153931.20470258663, "data_wait_s": 1.785199992809794e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 802200, "loss": 0.5233058333396912, "lr": 3e-05, "grad_norm": 0.26054259389607703, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117418832960, "tokens_per_second": 151372.29255512965, "data_wait_s": 1.4666999959445093e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 802300, "loss": 0.524161159992218, "lr": 3e-05, "grad_norm": 0.26103738770913526, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117433410377, "tokens_per_second": 151683.78917434896, "data_wait_s": 1.4295000255515333e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 802400, "loss": 0.533500075340271, "lr": 3e-05, "grad_norm": 0.2687727625425812, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117448067975, "tokens_per_second": 152703.83010156357, "data_wait_s": 2.6748999971459853e-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:af85851ccfb82768d845d8e8cdae225fe086f7b37eab42c3a7803f775f469d8e
-size 48879
+oid sha256:14fdeaabd666ea55628426220d9033c48356660ec7a736d9624d680c800533d8
+size 50997