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": 75577.29569366027,
-      "required_pages_per_second_upper_estimate": 0.006550848441821762
+      "required_token_positions_per_second": 75428.42164138392,
+      "required_pages_per_second_upper_estimate": 0.006537944416288284
     },
     {
       "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": 9447.161961707534,
-      "required_pages_per_second_upper_estimate": 0.003768614154183634
+      "required_token_positions_per_second": 9428.55270517299,
+      "required_pages_per_second_upper_estimate": 0.003761190643518825
     },
     {
       "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": 18894.323923415068,
-      "required_pages_per_second_upper_estimate": 0.004370869310938291
+      "required_token_positions_per_second": 18857.10541034598,
+      "required_pages_per_second_upper_estimate": 0.004362259462963215
     },
     {
       "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": 28341.4858851226,
-      "required_pages_per_second_upper_estimate": 0.0017258477941193816
+      "required_token_positions_per_second": 28285.65811551897,
+      "required_pages_per_second_upper_estimate": 0.0017224481758561906
     },
     {
       "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": 28341.4858851226,
-      "required_pages_per_second_upper_estimate": 0.0021818060376909637
+      "required_token_positions_per_second": 28285.65811551897,
+      "required_pages_per_second_upper_estimate": 0.0021775082614457187
     },
     {
       "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": 28341.4858851226,
-      "required_pages_per_second_upper_estimate": 0.3519064018416703
+      "required_token_positions_per_second": 28285.65811551897,
+      "required_pages_per_second_upper_estimate": 0.3512132077867188
     }
   ],
   "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": 0.05369300515624076,
-    "producer_p95_seconds": 0.13150947299982363,
-    "gpu_update_seconds": 0.975531068203921,
-    "producer_batches_per_second": 18.62439990255918,
-    "gpu_batches_per_second": 1.0250826781366682,
-    "producer_headroom": 18.16868073160056,
+    "producer_batch_seconds": 0.0493845022499908,
+    "producer_p95_seconds": 0.1278226449999238,
+    "gpu_update_seconds": 0.9774564865022844,
+    "producer_batches_per_second": 20.249267572605458,
+    "gpu_batches_per_second": 1.023063444571722,
+    "producer_headroom": 19.79277793576357,
     "sustainable": true,
-    "buffer_seconds": 60.4829262286431,
+    "buffer_seconds": 60.60230216314164,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 61,
+    "upload_pages_per_commit": 62,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790606918076636751-3952f1c5/train.jsonl

@@ -10,3 +10,4 @@
 {"step": 196000, "loss": 0.5849446766078472, "lr": 0.0003, "grad_norm": 0.17091316453886277, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28686768471, "tokens_per_second": 148838.12571125917, "data_wait_s": 1.9231000123909325e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 196100, "loss": 0.5884930640459061, "lr": 0.0003, "grad_norm": 0.13476272340182135, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28701200508, "tokens_per_second": 147894.88826712858, "data_wait_s": 1.4421000059883227e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 196200, "loss": 0.5501266270875931, "lr": 0.0003, "grad_norm": 0.12738395066953273, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28715614316, "tokens_per_second": 147553.1400489183, "data_wait_s": 1.3674999991053483e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 196300, "loss": 0.6402901001274586, "lr": 0.0003, "grad_norm": 0.17887708719374307, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28730226039, "tokens_per_second": 149410.7832142626, "data_wait_s": 1.371100006508641e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790606918.0a7f8bdc-23d3-4ecb-bd0c-4a4bf460026e-rl2cx.1192.0.1790606918076636751-3952f1c5

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:786a3c5f7d31f351fee146b87c50f3b59924c9bef81eba93b53f559efe1c43cf
-size 25581
+oid sha256:c6f78f35930a4407996f1ac4309df5dec86e053fec91d707c2937b30ab600d09
+size 27699