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": 77219.87608519998,
-      "required_pages_per_second_upper_estimate": 0.005342250202372122
+      "required_token_positions_per_second": 77159.44534043832,
+      "required_pages_per_second_upper_estimate": 0.005338069463230864
     },
     {
       "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": 9652.484510649998,
-      "required_pages_per_second_upper_estimate": 0.0040412936978116475
+      "required_token_positions_per_second": 9644.93066755479,
+      "required_pages_per_second_upper_estimate": 0.004038131061449865
     },
     {
       "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": 19304.969021299996,
-      "required_pages_per_second_upper_estimate": 0.004020715940867784
+      "required_token_positions_per_second": 19289.86133510958,
+      "required_pages_per_second_upper_estimate": 0.004017569408250797
     },
     {
       "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": 28957.453531949996,
-      "required_pages_per_second_upper_estimate": 0.0017359551237690178
+      "required_token_positions_per_second": 28934.792002664373,
+      "required_pages_per_second_upper_estimate": 0.0017345965996904964
     },
     {
       "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": 28957.453531949996,
-      "required_pages_per_second_upper_estimate": 0.0022513146417319414
+      "required_token_positions_per_second": 28934.792002664373,
+      "required_pages_per_second_upper_estimate": 0.0022495528075075176
     },
     {
       "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": 28957.453531949996,
-      "required_pages_per_second_upper_estimate": 0.37143512181667754
+      "required_token_positions_per_second": 28934.792002664373,
+      "required_pages_per_second_upper_estimate": 0.37114444405105596
     }
   ],
   "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": 720750,
-    "producer_batch_seconds": 0.051193315132818995,
-    "producer_p95_seconds": 0.11797608999995646,
-    "gpu_update_seconds": 0.9547800869125036,
-    "producer_batches_per_second": 19.533800407446563,
-    "gpu_batches_per_second": 1.0473616005479598,
-    "producer_headroom": 18.650483650753326,
+    "producer_batch_seconds": 0.04144186439063269,
+    "producer_p95_seconds": 0.11381703500001095,
+    "gpu_update_seconds": 0.9555278640832848,
+    "producer_batches_per_second": 24.13018851116252,
+    "gpu_batches_per_second": 1.046541956114886,
+    "producer_headroom": 23.05706748799814,
     "sustainable": true,
-    "buffer_seconds": 60.15114547548772,
+    "buffer_seconds": 60.19825543724694,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 45,
+    "upload_pages_per_commit": 52,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

metrics/void-byte-v1/1791254470898546162-7724d9f6/train.jsonl

@@ -3,3 +3,4 @@
 {"step": 760300, "loss": 0.5635800957679749, "lr": 3e-05, "grad_norm": 0.24654580489168332, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111285757983, "tokens_per_second": 153699.24148221326, "data_wait_s": 8.567999998376763e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 760400, "loss": 0.5628456473350525, "lr": 3e-05, "grad_norm": 0.24173337303194647, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111300311346, "tokens_per_second": 152641.78806016187, "data_wait_s": 7.332999985010247e-06, "gpu_peak_gb": 89.37899398803711}
 {"step": 760500, "loss": 0.5507456064224243, "lr": 3e-05, "grad_norm": 0.23829818174644574, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111314904726, "tokens_per_second": 152748.8207374359, "data_wait_s": 1.2248000075487653e-05, "gpu_peak_gb": 89.37899398803711}
+{"step": 760600, "loss": 0.5533353686332703, "lr": 3e-05, "grad_norm": 0.2419612001940234, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111329597564, "tokens_per_second": 153855.5123941884, "data_wait_s": 9.224999985235627e-06, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1791254470.1e0c9aa3-e42e-441e-9e17-2e25895f438d-9bhv5.366.0.1791254470898546162-7724d9f6

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3c759871dc1d3891f88e76bbd597447e3f9cde762dd14b41813e1bac947bc94e
-size 10755
+oid sha256:d65d59099c215c7501d59a91312f63037adfa0683b245648c5337eb0deb34ece
+size 12873