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": 78276.12513399479,
-      "required_pages_per_second_upper_estimate": 0.005415323962403199
+      "required_token_positions_per_second": 77878.63558719619,
+      "required_pages_per_second_upper_estimate": 0.005387824713252855
     },
     {
       "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": 9784.515641749349,
-      "required_pages_per_second_upper_estimate": 0.004096572375279405
+      "required_token_positions_per_second": 9734.829448399523,
+      "required_pages_per_second_upper_estimate": 0.004075769803689536
     },
     {
       "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": 19569.031283498698,
-      "required_pages_per_second_upper_estimate": 0.004075713146293625
+      "required_token_positions_per_second": 19469.658896799046,
+      "required_pages_per_second_upper_estimate": 0.0040550164987701535
     },
     {
       "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": 29353.546925248047,
-      "required_pages_per_second_upper_estimate": 0.0017597003178976434
+      "required_token_positions_per_second": 29204.48834519857,
+      "required_pages_per_second_upper_estimate": 0.0017507644836229506
     },
     {
       "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": 29353.546925248047,
-      "required_pages_per_second_upper_estimate": 0.0022821091608301526
+      "required_token_positions_per_second": 29204.48834519857,
+      "required_pages_per_second_upper_estimate": 0.0022705205118707093
     },
     {
       "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": 29353.546925248047,
-      "required_pages_per_second_upper_estimate": 0.37651578257395424
+      "required_token_positions_per_second": 29204.48834519857,
+      "required_pages_per_second_upper_estimate": 0.3746038191557134
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,18 +158,18 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 1.5711270929999728,
-    "producer_p95_seconds": 3.1246708119999766,
-    "gpu_update_seconds": 0.9418963939999685,
-    "producer_batches_per_second": 0.6364857460961736,
-    "gpu_batches_per_second": 1.0616878951550943,
-    "producer_headroom": 0.5995036290803655,
-    "sustainable": false,
-    "buffer_seconds": 60.281369215997984,
-    "buffer_drain_seconds": 150.51664282894168,
+    "producer_batch_seconds": 0.058775333203120894,
+    "producer_p95_seconds": 0.14470582400008425,
+    "gpu_update_seconds": 0.9467037967999714,
+    "producer_batches_per_second": 17.01394012593877,
+    "gpu_batches_per_second": 1.0562965981336288,
+    "producer_headroom": 16.10716171575362,
+    "sustainable": true,
+    "buffer_seconds": 60.58904299519817,
+    "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 7,
-    "max_pending_pages": 1661,
+    "upload_pages_per_commit": 16,
+    "max_pending_pages": 3920,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,
     "upload_sustainable": null,

metrics/void-byte-v1/1791033297667653750-ba4975eb/train.jsonl

@@ -0,0 +1 @@
+{"step": 600100, "loss": 0.6710267066955566, "lr": 0.0003, "grad_norm": 0.16837818088735748, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 87836984173, "tokens_per_second": 153948.01560736474, "data_wait_s": 1.1639000035756908e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791033297.83ac8725-d818-41af-bbb1-78dcd88792cf-7zwhs.737.0.1791033297667653750-ba4975eb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d6dc575f97faa63d8273c8f7e5461c57163e4e2e2eb8a0677430881ffec952d
-size 165
+oid sha256:b2d1b908b7fe0bc494c47a870ab5d80ccdf13f3d907a9198bc8a0020032a8ef7
+size 2283