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": 75547.34094335152,
-      "required_pages_per_second_upper_estimate": 0.00654825203998462
+      "required_token_positions_per_second": 75394.2626326947,
+      "required_pages_per_second_upper_estimate": 0.006534983599990319
     },
     {
       "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": 9443.41761791894,
-      "required_pages_per_second_upper_estimate": 0.003767120479463435
+      "required_token_positions_per_second": 9424.282829086838,
+      "required_pages_per_second_upper_estimate": 0.003759487326107722
     },
     {
       "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": 18886.83523583788,
-      "required_pages_per_second_upper_estimate": 0.004369136934863708
+      "required_token_positions_per_second": 18848.565658173677,
+      "required_pages_per_second_upper_estimate": 0.004360283941592658
     },
     {
       "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": 28330.25285375682,
-      "required_pages_per_second_upper_estimate": 0.0017251637614443664
+      "required_token_positions_per_second": 28272.848487260515,
+      "required_pages_per_second_upper_estimate": 0.001721668136702176
     },
     {
       "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": 28330.25285375682,
-      "required_pages_per_second_upper_estimate": 0.002180941287841405
+      "required_token_positions_per_second": 28272.848487260515,
+      "required_pages_per_second_upper_estimate": 0.002176522140803062
     },
     {
       "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": 28330.25285375682,
-      "required_pages_per_second_upper_estimate": 0.35176692518664493
+      "required_token_positions_per_second": 28272.848487260515,
+      "required_pages_per_second_upper_estimate": 0.35105415507481674
     }
   ],
   "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": 0.05406646315624819,
-    "producer_p95_seconds": 0.13742285300008916,
-    "gpu_update_seconds": 0.975917869237572,
-    "producer_batches_per_second": 18.495753959530735,
-    "gpu_batches_per_second": 1.0246763908332184,
-    "producer_headroom": 18.050336794127617,
+    "producer_batch_seconds": 0.049242815632805303,
+    "producer_p95_seconds": 0.13248381600010362,
+    "gpu_update_seconds": 0.9778993443995547,
+    "producer_batches_per_second": 20.307530898655706,
+    "gpu_batches_per_second": 1.0226001333644572,
+    "producer_headroom": 19.858721152169117,
     "sustainable": true,
-    "buffer_seconds": 60.50690789272946,
+    "buffer_seconds": 60.629759352772396,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 65,
+    "upload_pages_per_commit": 66,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -12,3 +12,4 @@
 {"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}
 {"step": 196400, "loss": 0.6060016378760338, "lr": 0.0003, "grad_norm": 0.14864991768030777, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28744944508, "tokens_per_second": 150125.7256151432, "data_wait_s": 1.5746000144645222e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 196500, "loss": 0.5726331695914268, "lr": 0.0003, "grad_norm": 0.13443351313904545, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28759668846, "tokens_per_second": 150724.4681650211, "data_wait_s": 1.4403000022866763e-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:c287a42bb9ff4508a743d77bb8fc9416db4c4ddd19fbafad9f256b590bff72cd
-size 29817
+oid sha256:317ec4e5b4ab0ff4a934aa667ea4034dbf33958440edceb55a54aba8d6f25470
+size 31935