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": 77377.4613779064,
-      "required_pages_per_second_upper_estimate": 0.005353152318569808
+      "required_token_positions_per_second": 77299.80957603874,
+      "required_pages_per_second_upper_estimate": 0.0053477801867395985
     },
     {
       "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": 9672.1826722383,
-      "required_pages_per_second_upper_estimate": 0.00404954090672428
+      "required_token_positions_per_second": 9662.476197004842,
+      "required_pages_per_second_upper_estimate": 0.004045477008238283
     },
     {
       "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": 19344.3653444766,
-      "required_pages_per_second_upper_estimate": 0.0040289211560345616
+      "required_token_positions_per_second": 19324.952394009684,
+      "required_pages_per_second_upper_estimate": 0.004024877950406234
     },
     {
       "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": 29016.548016714896,
-      "required_pages_per_second_upper_estimate": 0.00173949774789862
+      "required_token_positions_per_second": 28987.428591014526,
+      "required_pages_per_second_upper_estimate": 0.0017377520827906714
     },
     {
       "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": 29016.548016714896,
-      "required_pages_per_second_upper_estimate": 0.002255908977993244
+      "required_token_positions_per_second": 28987.428591014526,
+      "required_pages_per_second_upper_estimate": 0.002253645070727858
     },
     {
       "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": 29016.548016714896,
-      "required_pages_per_second_upper_estimate": 0.37219312241652747
+      "required_token_positions_per_second": 28987.428591014526,
+      "required_pages_per_second_upper_estimate": 0.37181960968964645
     }
   ],
   "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": 0.04802520735156879,
-    "producer_p95_seconds": 0.11499261699998442,
-    "gpu_update_seconds": 0.9528356020872453,
-    "producer_batches_per_second": 20.82239838507088,
-    "gpu_batches_per_second": 1.0494989878730792,
-    "producer_headroom": 19.840322502139493,
+    "producer_batch_seconds": 0.04329927381250531,
+    "producer_p95_seconds": 0.11905074000003424,
+    "gpu_update_seconds": 0.9537927765200354,
+    "producer_batches_per_second": 23.095075550925035,
+    "gpu_batches_per_second": 1.0484457679041712,
+    "producer_headroom": 22.02791623365677,
     "sustainable": true,
-    "buffer_seconds": 60.02864293149645,
+    "buffer_seconds": 60.08894492076223,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 49,
+    "upload_pages_per_commit": 60,
     "max_pending_pages": 4096,
     "generation_bytes_per_second": 0,
     "upload_bytes_per_second": null,

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

@@ -2,3 +2,4 @@
 {"step": 925200, "loss": 0.5638423562049866, "lr": 3e-05, "grad_norm": 0.3383034160205606, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135422770685, "tokens_per_second": 154311.20983017658, "data_wait_s": 1.4731999954165076e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 925300, "loss": 0.5204002857208252, "lr": 3e-05, "grad_norm": 0.3165123073884374, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135437541266, "tokens_per_second": 154976.86957509432, "data_wait_s": 1.3204000083533174e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 925400, "loss": 0.5438966155052185, "lr": 3e-05, "grad_norm": 0.335793158948784, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135452275707, "tokens_per_second": 154698.59183631587, "data_wait_s": 7.487999937438872e-06, "gpu_peak_gb": 89.37899398803711}
+{"step": 925500, "loss": 0.4629822075366974, "lr": 3e-05, "grad_norm": 0.3368484758242998, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135466831988, "tokens_per_second": 152979.93765096815, "data_wait_s": 1.1587999892981315e-05, "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 sha256:9cd0ccd6e805e64828c82faefe5f7ec81c4bb3ab09afc50f09381a6314edc654
-size 8637
+oid sha256:65d47019d6c81aae40e200b6b084f33ecf687ec9697d6ef2e975d32f08007d5a
+size 10755