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": 77477.60150821909,
-      "required_pages_per_second_upper_estimate": 0.005360080245141954
+      "required_token_positions_per_second": 77544.88908671791,
+      "required_pages_per_second_upper_estimate": 0.005364735355951195
     },
     {
       "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": 9684.700188527386,
-      "required_pages_per_second_upper_estimate": 0.004054781729399055
+      "required_token_positions_per_second": 9693.11113583974,
+      "required_pages_per_second_upper_estimate": 0.0040583032173981855
     },
     {
       "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": 19369.40037705477,
-      "required_pages_per_second_upper_estimate": 0.004034135293102388
+      "required_token_positions_per_second": 19386.22227167948,
+      "required_pages_per_second_upper_estimate": 0.004037638850129521
     },
     {
       "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": 29054.100565582157,
-      "required_pages_per_second_upper_estimate": 0.0017417489658637902
+      "required_token_positions_per_second": 29079.333407519218,
+      "required_pages_per_second_upper_estimate": 0.001743261636209598
     },
     {
       "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": 29054.100565582157,
-      "required_pages_per_second_upper_estimate": 0.0022588285234914677
+      "required_token_positions_per_second": 29079.333407519218,
+      "required_pages_per_second_upper_estimate": 0.002260790265964532
     },
     {
       "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": 29054.100565582157,
-      "required_pages_per_second_upper_estimate": 0.3726748061926111
+      "required_token_positions_per_second": 29079.333407519218,
+      "required_pages_per_second_upper_estimate": 0.37299846599606495
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,17 +158,17 @@
     "prefetch_batches": 64,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.05401758545312063,
-    "producer_p95_seconds": 0.1506637600000431,
-    "gpu_update_seconds": 0.9516040580086709,
-    "producer_batches_per_second": 18.512489805155283,
-    "gpu_batches_per_second": 1.0508572253176416,
-    "producer_headroom": 17.616560422429917,
+    "producer_batch_seconds": 0.05146706214842922,
+    "producer_p95_seconds": 0.13528440500022043,
+    "gpu_update_seconds": 0.9507783281184462,
+    "producer_batches_per_second": 19.429902509609636,
+    "gpu_batches_per_second": 1.0517698715103883,
+    "producer_headroom": 18.47353022359105,
     "sustainable": true,
-    "buffer_seconds": 60.902659712554936,
+    "buffer_seconds": 60.84981299958056,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 84,
+    "upload_pages_per_commit": 83,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791463982175382982-d4a3b372/train.jsonl

@@ -37,3 +37,4 @@
 {"step": 928700, "loss": 0.5363717675209045, "lr": 3e-05, "grad_norm": 0.3355081071269844, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135935459183, "tokens_per_second": 153776.51483855143, "data_wait_s": 6.9359998633444775e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 928800, "loss": 0.5778369307518005, "lr": 3e-05, "grad_norm": 0.3463705909920343, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135949889323, "tokens_per_second": 151523.70106611072, "data_wait_s": 1.3439999747788534e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 928900, "loss": 0.5194796323776245, "lr": 3e-05, "grad_norm": 0.3417008088336159, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135964381838, "tokens_per_second": 152211.75039808516, "data_wait_s": 7.288000233529601e-06, "gpu_peak_gb": 89.37899398803711}
+{"step": 929000, "loss": 0.5087798237800598, "lr": 3e-05, "grad_norm": 0.3377837229940126, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135978948262, "tokens_per_second": 152233.88213437382, "data_wait_s": 6.504000339191407e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791463982.2e076f0c-3fe2-4316-8f17-4e30b2e706e2-b5972.360.0.1791463982175382982-d4a3b372

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f4ad52c504a6e3aebd5e25d7eeea4a318de74ff4508439821ff8ac3365136b69
-size 82767
+oid sha256:8f43e196fb200beeb05d912d8cb6ab76bedc158767e77b8ebd0088da8aaaba1a
+size 84885