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": 76393.02712426,
-      "required_pages_per_second_upper_estimate": 0.006621553974773737
+      "required_token_positions_per_second": 76216.36192086617,
+      "required_pages_per_second_upper_estimate": 0.006606241082697424
     },
     {
       "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": 9549.1283905325,
-      "required_pages_per_second_upper_estimate": 0.0038092900871758814
+      "required_token_positions_per_second": 9527.045240108271,
+      "required_pages_per_second_upper_estimate": 0.003800480788299135
     },
     {
       "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": 19098.256781065,
-      "required_pages_per_second_upper_estimate": 0.004418045588459891
+      "required_token_positions_per_second": 19054.090480216542,
+      "required_pages_per_second_upper_estimate": 0.0044078284920589465
     },
     {
       "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": 28647.385171597496,
-      "required_pages_per_second_upper_estimate": 0.001744475455736182
+      "required_token_positions_per_second": 28581.13572032481,
+      "required_pages_per_second_upper_estimate": 0.0017404412117376837
     },
     {
       "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": 28647.385171597496,
-      "required_pages_per_second_upper_estimate": 0.0022053550115472224
+      "required_token_positions_per_second": 28581.13572032481,
+      "required_pages_per_second_upper_estimate": 0.0022002549453980324
     },
     {
       "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": 28647.385171597496,
-      "required_pages_per_second_upper_estimate": 0.35570464720063444
+      "required_token_positions_per_second": 28581.13572032481,
+      "required_pages_per_second_upper_estimate": 0.3548820507384781
     }
   ],
   "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": 688946,
-    "producer_batch_seconds": 0.055733439015615005,
-    "producer_p95_seconds": 0.13183292100006838,
-    "gpu_update_seconds": 0.965114262065763,
-    "producer_batches_per_second": 17.942549709158033,
-    "gpu_batches_per_second": 1.036146743764377,
-    "producer_headroom": 17.316610622132327,
+    "producer_batch_seconds": 0.05273894703123716,
+    "producer_p95_seconds": 0.138644443999965,
+    "gpu_update_seconds": 0.9673513421770278,
+    "producer_batches_per_second": 18.961319030653044,
+    "gpu_batches_per_second": 1.033750568588137,
+    "producer_headroom": 18.342257413749042,
     "sustainable": true,
-    "buffer_seconds": 60.80219851014307,
+    "buffer_seconds": 60.94313455715275,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 62,
+    "upload_pages_per_commit": 67,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1790519810984249790-cb9c9513/train.jsonl

@@ -8,3 +8,4 @@
 {"step": 115800, "loss": 0.603627547621727, "lr": 0.0003, "grad_norm": 0.15810096959168768, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16945791532, "tokens_per_second": 152467.27985714795, "data_wait_s": 9.756999816090683e-06, "gpu_peak_gb": 85.46427917480469}
 {"step": 115900, "loss": 0.5959490407258272, "lr": 0.0003, "grad_norm": 0.14769660827695635, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16960422649, "tokens_per_second": 151080.4202959854, "data_wait_s": 1.5132000044104643e-05, "gpu_peak_gb": 85.46427917480469}
 {"step": 116000, "loss": 0.6013966966420412, "lr": 0.0003, "grad_norm": 0.12939835376412753, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16975010873, "tokens_per_second": 149040.05846871884, "data_wait_s": 1.3589999980467837e-05, "gpu_peak_gb": 85.46427917480469}
+{"step": 116100, "loss": 0.5669870059937239, "lr": 0.0003, "grad_norm": 0.1382701306324106, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16989562756, "tokens_per_second": 149802.0055713855, "data_wait_s": 1.2924999964525341e-05, "gpu_peak_gb": 85.46427917480469}

runs/void-byte-v1/events.out.tfevents.1790519810.cad0b19a-59f2-43f0-8fb4-2a94097843e2-tjh7c.1336.0.1790519810984249790-cb9c9513

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:38ca3b145c03612fb6dc8c16882fe12875d2464e9905e693760022b60b0aae2a
-size 21345
+oid sha256:497eb05158f4678b209201fabbccb88dcadcd96b9a7736868d1030b0a163a7bc
+size 23463