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": 75800.36564116608,
-      "required_pages_per_second_upper_estimate": 0.006570183579505952
+      "required_token_positions_per_second": 76098.16996917009,
+      "required_pages_per_second_upper_estimate": 0.006595996503879172
     },
     {
       "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": 9475.04570514576,
-      "required_pages_per_second_upper_estimate": 0.0037797373963402584
+      "required_token_positions_per_second": 9512.271246146262,
+      "required_pages_per_second_upper_estimate": 0.0037945872212167948
     },
     {
       "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": 18950.09141029152,
-      "required_pages_per_second_upper_estimate": 0.00438377013755291
+      "required_token_positions_per_second": 19024.542492292523,
+      "required_pages_per_second_upper_estimate": 0.004400993085079552
     },
     {
       "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": 28425.137115437283,
-      "required_pages_per_second_upper_estimate": 0.0017309417151614589
+      "required_token_positions_per_second": 28536.81373843878,
+      "required_pages_per_second_upper_estimate": 0.0017377422355802918
     },
     {
       "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": 28425.137115437283,
-      "required_pages_per_second_upper_estimate": 0.0021882457409620137
+      "required_token_positions_per_second": 28536.81373843878,
+      "required_pages_per_second_upper_estimate": 0.0021968429165413477
     },
     {
       "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": 28425.137115437283,
-      "required_pages_per_second_upper_estimate": 0.35294507015952026
+      "required_token_positions_per_second": 28536.81373843878,
+      "required_pages_per_second_upper_estimate": 0.3543317200595848
     }
   ],
   "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": 0.04761540457031366,
-    "producer_p95_seconds": 0.11974704199997177,
-    "gpu_update_seconds": 0.9726602157702441,
-    "producer_batches_per_second": 21.00160670741126,
-    "gpu_batches_per_second": 1.028108257936823,
-    "producer_headroom": 20.42742731155244,
+    "producer_batch_seconds": 0.05150888622656602,
+    "producer_p95_seconds": 0.11851246200001242,
+    "gpu_update_seconds": 0.9688537849184767,
+    "producer_batches_per_second": 19.41412585784555,
+    "gpu_batches_per_second": 1.0321474876460786,
+    "producer_headroom": 18.809449318257332,
     "sustainable": true,
-    "buffer_seconds": 60.304933377755134,
+    "buffer_seconds": 60.06893466494556,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 40,
+    "upload_pages_per_commit": 47,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790563279934888633-82973089/train.jsonl

@@ -1,3 +1,4 @@
 {"step": 155100, "loss": 0.5673591010272503, "lr": 0.0003, "grad_norm": 0.13793740573041358, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22698961668, "tokens_per_second": 44241.941022346604, "data_wait_s": 1.6559999949095072e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 155200, "loss": 0.6261652167886496, "lr": 0.0003, "grad_norm": 0.146581492453678, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22713619019, "tokens_per_second": 150427.58678988548, "data_wait_s": 1.6635999941172486e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 155300, "loss": 0.6224912591278553, "lr": 0.0003, "grad_norm": 0.16059344946226572, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22728190005, "tokens_per_second": 150311.51459487772, "data_wait_s": 1.39969999963796e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 155400, "loss": 0.6191136613488197, "lr": 0.0003, "grad_norm": 0.1460772445792664, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 22742793235, "tokens_per_second": 150671.66686189847, "data_wait_s": 1.2129999959142879e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790563279.bdfa290c-aca9-4e96-a5d2-8425921a3278-d85hq.1304.0.1790563279934888633-82973089

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e8cdcf47b79c42dc1766ea20f834313ef8e5edda16a621c8358b67b387f51fbb
-size 6519
+oid sha256:6838fff3a026b6d9f81d3124de23cc55cbe23412d5d0b49e5b9c10b50f37f00e
+size 8637