Ready Cohorts: Bounding GPU Opportunity and Avoiding Host Round Trips in LLM-Agent Control
Paper • 2608.12123 • Published • 1
Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code: DatasetGenerationCastError
Exception: DatasetGenerationCastError
Message: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 9 new columns ({'technical_repetitions_paired', 'cuda_host_launch_over_cuda_host_graph_paired_wall_ratio_p95', 'cuda_device_graph_over_cuda_host_graph_paired_wall_ratio_p99', 'cuda_host_launch_over_cuda_host_graph_paired_wall_ratio_median', 'cuda_device_graph_over_cuda_host_graph_wall_ratio_of_medians', 'cuda_device_graph_over_cuda_host_graph_paired_wall_ratio_p95', 'cuda_device_graph_over_cuda_host_graph_paired_wall_ratio_median', 'cuda_host_launch_over_cuda_host_graph_wall_ratio_of_medians', 'cuda_host_launch_over_cuda_host_graph_paired_wall_ratio_p99'}) and 13 missing columns ({'device_ns_p95', 'status_ok_rows', 'technical_repetitions', 'checksum_equal_rows', 'mechanism', 'wall_ns_p99', 'wall_ns_p95', 'device_ns_p99', 'device_ns_rows', 'device_time_scope', 'device_ns_median', 'wall_ns_median', 'field_exact_rows'}).
This happened while the csv dataset builder was generating data using
hf://datasets/josefchen/ready-cohorts/native/contrasts.csv (at revision 41ff8d06578d4ec68934040a8fb925d3c8b01767), ['hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/native/cell-summary.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/native/contrasts.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/paper/claim-evidence-map.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/resident/cell-summary.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/resident/contrasts.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/source/session-summary.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/trace/cell-seed-evaluations.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/trace/summary.csv']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1837, in _prepare_split_single
writer.write_table(table)
~~~~~~~~~~~~~~~~~~^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
self._write_table(pa_table, writer_batch_size=writer_batch_size)
~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
pa_table = table_cast(pa_table, self._schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
placement_id: string
provider: string
requested_gpu: string
device_name: string
device_uuid: string
source_file: string
source_sha256: string
agents: int64
steps: int64
analysis_unit: string
technical_repetitions_paired: int64
cuda_device_graph_over_cuda_host_graph_wall_ratio_of_medians: double
cuda_device_graph_over_cuda_host_graph_paired_wall_ratio_median: double
cuda_device_graph_over_cuda_host_graph_paired_wall_ratio_p95: double
cuda_device_graph_over_cuda_host_graph_paired_wall_ratio_p99: double
cuda_host_launch_over_cuda_host_graph_wall_ratio_of_medians: double
cuda_host_launch_over_cuda_host_graph_paired_wall_ratio_median: double
cuda_host_launch_over_cuda_host_graph_paired_wall_ratio_p95: double
cuda_host_launch_over_cuda_host_graph_paired_wall_ratio_p99: double
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 3403
to
{'placement_id': Value('string'), 'provider': Value('string'), 'requested_gpu': Value('string'), 'device_name': Value('string'), 'device_uuid': Value('string'), 'source_file': Value('string'), 'source_sha256': Value('string'), 'agents': Value('int64'), 'steps': Value('int64'), 'mechanism': Value('string'), 'analysis_unit': Value('string'), 'technical_repetitions': Value('int64'), 'status_ok_rows': Value('int64'), 'field_exact_rows': Value('int64'), 'checksum_equal_rows': Value('int64'), 'wall_ns_median': Value('float64'), 'wall_ns_p95': Value('float64'), 'wall_ns_p99': Value('float64'), 'device_time_scope': Value('string'), 'device_ns_rows': Value('int64'), 'device_ns_median': Value('float64'), 'device_ns_p95': Value('float64'), 'device_ns_p99': Value('float64')}
because column names don't match
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1683, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1839, in _prepare_split_single
raise DatasetGenerationCastError.from_cast_error(
...<4 lines>...
)
datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 9 new columns ({'technical_repetitions_paired', 'cuda_host_launch_over_cuda_host_graph_paired_wall_ratio_p95', 'cuda_device_graph_over_cuda_host_graph_paired_wall_ratio_p99', 'cuda_host_launch_over_cuda_host_graph_paired_wall_ratio_median', 'cuda_device_graph_over_cuda_host_graph_wall_ratio_of_medians', 'cuda_device_graph_over_cuda_host_graph_paired_wall_ratio_p95', 'cuda_device_graph_over_cuda_host_graph_paired_wall_ratio_median', 'cuda_host_launch_over_cuda_host_graph_wall_ratio_of_medians', 'cuda_host_launch_over_cuda_host_graph_paired_wall_ratio_p99'}) and 13 missing columns ({'device_ns_p95', 'status_ok_rows', 'technical_repetitions', 'checksum_equal_rows', 'mechanism', 'wall_ns_p99', 'wall_ns_p95', 'device_ns_p99', 'device_ns_rows', 'device_time_scope', 'device_ns_median', 'wall_ns_median', 'field_exact_rows'}).
This happened while the csv dataset builder was generating data using
hf://datasets/josefchen/ready-cohorts/native/contrasts.csv (at revision 41ff8d06578d4ec68934040a8fb925d3c8b01767), ['hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/native/cell-summary.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/native/contrasts.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/paper/claim-evidence-map.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/resident/cell-summary.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/resident/contrasts.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/source/session-summary.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/trace/cell-seed-evaluations.csv', 'hf://datasets/josefchen/ready-cohorts@41ff8d06578d4ec68934040a8fb925d3c8b01767/trace/summary.csv']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
placement_id string | provider string | requested_gpu string | device_name string | device_uuid string | source_file string | source_sha256 string | agents int64 | steps int64 | mechanism string | analysis_unit string | technical_repetitions int64 | status_ok_rows int64 | field_exact_rows int64 | checksum_equal_rows int64 | wall_ns_median float64 | wall_ns_p95 float64 | wall_ns_p99 float64 | device_time_scope string | device_ns_rows int64 | device_ns_median float64 | device_ns_p95 float64 | device_ns_p99 float64 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 1 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 174 | 178 | 178 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 1 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 16,211 | 16,938.4 | 16,973.67 | complete_parent_and_child_execution_environment | 50 | 11,056 | 12,003.2 | 12,064 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 1 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 9,140 | 9,394.7 | 9,427.99 | transition_kernels | 50 | 3,600 | 3,793.6 | 3,840 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 1 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 10,113 | 10,795.8 | 18,012.52 | transition_kernels | 50 | 5,136 | 5,571.2 | 10,344.64 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 8 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 1,240.5 | 1,253 | 1,254.51 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 8 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 23,510.5 | 24,175.75 | 24,453.42 | complete_parent_and_child_execution_environment | 50 | 18,464 | 19,232 | 19,506.24 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 8 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 16,098.5 | 16,393.35 | 19,171.81 | transition_kernels | 50 | 10,528 | 10,784 | 10,881.28 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 8 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 28,065.5 | 29,239.35 | 29,332.16 | transition_kernels | 50 | 23,136 | 24,280 | 24,368.32 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 64 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 9,766 | 9,831.9 | 12,580.38 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 64 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 77,915.5 | 78,820.25 | 81,152.78 | complete_parent_and_child_execution_environment | 50 | 72,848 | 73,763.2 | 76,059.2 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 64 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 70,657 | 71,022.5 | 71,095.51 | transition_kernels | 50 | 65,183 | 65,376 | 65,408 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 64 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 160,961.5 | 174,374.75 | 176,492.77 | transition_kernels | 50 | 156,000 | 169,438.4 | 171,511.68 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 1 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 1,122 | 1,135.4 | 1,145.02 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 1 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 16,340 | 17,044.45 | 17,287.17 | complete_parent_and_child_execution_environment | 50 | 11,280 | 12,049.6 | 12,291.2 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 1 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 9,142.5 | 9,298.55 | 9,373.54 | transition_kernels | 50 | 3,536 | 3,680 | 3,794.24 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 1 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 9,855 | 10,173.65 | 10,814.31 | transition_kernels | 50 | 4,864 | 5,216 | 5,847.36 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 8 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 9,552 | 9,589.2 | 9,609.57 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 8 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 23,332.5 | 24,342.65 | 25,311.97 | complete_parent_and_child_execution_environment | 50 | 18,304 | 19,412.8 | 20,367.36 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 8 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 16,370 | 16,586.5 | 17,039.54 | transition_kernels | 50 | 10,816 | 10,979.2 | 11,105.28 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 8 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 26,667.5 | 28,819.85 | 33,220.36 | transition_kernels | 50 | 21,712 | 23,913.6 | 28,295.04 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 64 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 77,712 | 82,949.75 | 83,635.56 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 64 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 79,419.5 | 80,179.75 | 80,387.6 | complete_parent_and_child_execution_environment | 50 | 74,336 | 75,185.6 | 75,200 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 64 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 72,067.5 | 72,385.55 | 74,638.77 | transition_kernels | 50 | 66,464 | 66,673.6 | 69,265.92 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 256 | 64 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 160,984 | 171,752 | 174,655.79 | transition_kernels | 50 | 156,032 | 166,726.4 | 169,701.44 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 1 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 8,670 | 8,701.1 | 8,713.61 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 1 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 16,494 | 17,333.05 | 23,096.47 | complete_parent_and_child_execution_environment | 50 | 11,568 | 12,344 | 13,117.12 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 1 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 9,316.5 | 9,607.55 | 9,650.89 | transition_kernels | 50 | 3,760 | 3,968 | 4,000.64 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 1 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 9,979 | 10,532.6 | 15,958.42 | transition_kernels | 50 | 5,008 | 5,612.8 | 10,939.52 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 8 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 75,529 | 78,715.9 | 81,037.83 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 8 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 24,201 | 24,992.6 | 25,802.18 | complete_parent_and_child_execution_environment | 50 | 19,200 | 20,003.2 | 20,160.64 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 8 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 16,868.5 | 17,067.25 | 17,167.18 | transition_kernels | 50 | 11,296 | 11,520 | 11,664.32 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 8 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 27,168 | 32,328.8 | 35,436.55 | transition_kernels | 50 | 22,256 | 27,112 | 30,537.92 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 64 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 625,035 | 631,964.35 | 632,428.89 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 64 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 83,643 | 84,719.1 | 85,225.37 | complete_parent_and_child_execution_environment | 50 | 78,592 | 79,601.6 | 79,936.64 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 64 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 75,884 | 76,568 | 79,040.35 | transition_kernels | 50 | 70,784 | 71,043.2 | 71,088.32 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 2,048 | 64 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 164,041.5 | 173,596.7 | 181,302.24 | transition_kernels | 50 | 159,056 | 168,593.6 | 176,295.68 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 1 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 68,827 | 75,610.95 | 78,253.51 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 1 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 16,751.5 | 17,344.65 | 17,699.34 | complete_parent_and_child_execution_environment | 50 | 11,808 | 12,401.6 | 12,771.84 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 1 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 8,979.5 | 9,419.35 | 10,741.44 | transition_kernels | 50 | 3,952 | 4,241.6 | 4,384 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 1 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 10,533.5 | 10,925.25 | 10,987.52 | transition_kernels | 50 | 5,520 | 5,937.6 | 6,064.96 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 8 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 610,964 | 625,975.4 | 635,107.47 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 8 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 25,912.5 | 26,864.3 | 33,641.74 | complete_parent_and_child_execution_environment | 50 | 20,944 | 21,955.2 | 28,677.44 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 8 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 17,517 | 18,232.5 | 19,063.84 | transition_kernels | 50 | 12,544 | 13,158.4 | 14,045.76 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 8 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 27,979 | 31,795.5 | 34,260.79 | transition_kernels | 50 | 22,960 | 24,542.4 | 28,837.12 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 64 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 5,011,056 | 5,038,903.35 | 5,044,558.81 | null | 0 | null | null | null |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 64 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 90,648.5 | 91,925.35 | 92,858.15 | complete_parent_and_child_execution_environment | 50 | 85,696 | 86,880 | 87,933.76 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 64 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 83,042.5 | 83,868 | 86,563.66 | transition_kernels | 50 | 78,080 | 78,819.2 | 78,992.96 |
native-dispatch-001-lambda-h100-p1 | lambda | H100 (80 GB SXM5) | NVIDIA H100 80GB HBM3 | 7964af6ebe675c48d8719797603d3893 | data/raw/native-dispatch-001-lambda-h100-p1-20260811T210728Z-p2682.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 16,384 | 64 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 164,254 | 174,003.6 | 177,913.34 | transition_kernels | 50 | 159,312 | 169,044.8 | 172,930.24 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 1 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 158 | 176 | 180.55 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 1 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 11,554 | 12,056.05 | 12,793.58 | complete_parent_and_child_execution_environment | 50 | 8,800 | 9,571.2 | 10,056.32 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 1 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 5,793.5 | 6,131.4 | 6,253.47 | transition_kernels | 50 | 3,248 | 3,936 | 4,048.96 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 1 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 6,347.5 | 6,650.75 | 8,008.15 | transition_kernels | 50 | 3,888 | 4,195.2 | 5,424 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 8 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 1,136 | 1,162.95 | 1,179.08 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 8 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 20,462 | 20,703 | 21,718.47 | complete_parent_and_child_execution_environment | 50 | 17,808 | 18,353.6 | 19,345.92 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 8 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 13,274.5 | 13,671.2 | 13,766.14 | transition_kernels | 50 | 10,624 | 11,537.6 | 11,600.32 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 8 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 17,814.5 | 18,114.15 | 18,226.75 | transition_kernels | 50 | 15,168 | 16,017.05 | 16,047.32 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 64 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 9,041.5 | 9,084.55 | 10,572.65 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 64 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 102,460.5 | 102,859.45 | 102,908.5 | complete_parent_and_child_execution_environment | 50 | 99,840 | 100,177.6 | 100,272.96 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 64 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 84,536 | 84,916.8 | 85,096.85 | transition_kernels | 50 | 81,920 | 82,659.2 | 82,880.64 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 32 | 64 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 109,717 | 111,012.05 | 115,587.24 | transition_kernels | 50 | 107,008 | 108,454.4 | 113,358.4 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 1 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 1,047 | 1,060.55 | 1,066 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 1 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 11,577.5 | 11,931.55 | 12,105.65 | complete_parent_and_child_execution_environment | 50 | 8,640 | 9,588.8 | 9,777.6 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 1 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 5,894.5 | 6,285.8 | 6,604.13 | transition_kernels | 50 | 3,344 | 3,921.6 | 4,016.32 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 1 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 6,410.5 | 6,749.35 | 8,149.19 | transition_kernels | 50 | 3,968 | 4,096 | 4,193.92 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 8 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 8,730.5 | 8,763.75 | 10,067.93 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 8 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 21,584 | 21,945.75 | 22,304.1 | complete_parent_and_child_execution_environment | 50 | 18,640 | 19,508.8 | 19,633.6 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 8 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 14,299 | 14,475.15 | 14,508.24 | transition_kernels | 50 | 11,840 | 12,192 | 12,288 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 8 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 18,609.5 | 18,872.4 | 18,939.15 | transition_kernels | 50 | 16,160 | 16,384 | 16,384 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 64 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 70,673.5 | 73,459.75 | 73,610.46 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 64 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 109,917.5 | 110,349.1 | 110,431.54 | complete_parent_and_child_execution_environment | 50 | 106,912 | 108,003.2 | 108,178.24 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 64 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 92,378.5 | 92,684.85 | 92,859.76 | transition_kernels | 50 | 89,856 | 90,112 | 90,160.32 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 256 | 64 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 115,785 | 116,617.85 | 117,092.34 | transition_kernels | 50 | 112,928 | 114,324.8 | 114,531.2 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 1 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 8,139 | 8,214.2 | 8,225.61 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 1 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 11,669.5 | 12,090.35 | 13,303.03 | complete_parent_and_child_execution_environment | 50 | 8,768 | 9,553.6 | 9,680.32 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 1 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 6,025.5 | 6,231.65 | 6,283.26 | transition_kernels | 50 | 3,536 | 4,032 | 4,096 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 1 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 6,618 | 6,872.6 | 25,207.37 | transition_kernels | 50 | 4,096 | 4,291.2 | 9,980.48 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 8 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 69,257 | 70,891.35 | 72,186.36 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 8 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 22,116.5 | 22,321.15 | 22,371.75 | complete_parent_and_child_execution_environment | 50 | 19,216 | 19,921.6 | 19,968.64 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 8 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 14,788 | 15,156.15 | 16,147.34 | transition_kernels | 50 | 12,272 | 12,384 | 12,497.6 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 8 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 19,422 | 19,816.1 | 21,303.9 | transition_kernels | 50 | 16,768 | 17,553.6 | 19,100.16 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 64 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 568,404.5 | 598,380.85 | 607,622.72 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 64 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 114,245 | 114,856.85 | 115,098.28 | complete_parent_and_child_execution_environment | 50 | 111,184 | 112,227.2 | 112,465.6 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 64 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 97,393 | 98,004.15 | 112,345.87 | transition_kernels | 50 | 94,656 | 95,488 | 104,691.84 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 2,048 | 64 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 121,389 | 122,407.4 | 126,764.42 | transition_kernels | 50 | 118,768 | 120,169.6 | 123,778.56 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 1 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 64,811.5 | 66,425.4 | 67,905.18 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 1 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 12,474.5 | 13,035.1 | 13,232.41 | complete_parent_and_child_execution_environment | 50 | 9,872 | 10,240 | 10,240 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 1 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 6,629 | 6,870.55 | 7,163.29 | transition_kernels | 50 | 4,096 | 4,529.6 | 4,880.32 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 1 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 7,283 | 7,951.85 | 10,050.12 | transition_kernels | 50 | 4,880 | 5,908.8 | 7,956.8 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 8 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 554,509.5 | 596,614.65 | 627,253.82 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 8 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 26,981 | 27,629.35 | 30,834.53 | complete_parent_and_child_execution_environment | 50 | 24,272 | 24,980.8 | 27,092.16 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 8 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 19,498 | 19,912.55 | 20,694.72 | transition_kernels | 50 | 16,976 | 17,827.2 | 18,114.56 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 8 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 24,384.5 | 24,966.4 | 26,320.55 | transition_kernels | 50 | 21,984 | 22,464 | 22,528.64 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 64 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 4,636,033.5 | 4,700,980.5 | 4,766,733.1 | null | 0 | null | null | null |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 64 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 155,674 | 158,022.45 | 161,918.09 | complete_parent_and_child_execution_environment | 50 | 153,168 | 154,872 | 155,652.48 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 64 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 134,836.5 | 136,454.3 | 136,707.58 | transition_kernels | 50 | 132,096 | 134,225.6 | 134,497.92 |
native-dispatch-001-local | local | GTX 1660 Ti | NVIDIA GeForce GTX 1660 Ti | 50798f9011b9a145d0237e476534cf0c | data/raw/native-dispatch-001-local-20260811T205817Z-p3867952.csv | a5c1f4a349075b6e76116c4f52163b488ccdae32a7bf66e95fed752e363d3ac6 | 16,384 | 64 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 159,578 | 161,477.85 | 168,011.59 | transition_kernels | 50 | 157,360 | 158,566.4 | 159,313.6 |
native-dispatch-001-modal-l4-p1 | modal | L4 | NVIDIA L4 | 7aa089db4b574bf77f4eca092af33188 | data/raw/native-dispatch-001-modal-l4-p1-20260811T205922Z-p117.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 1 | cpu_cpp | placement | 50 | 50 | 50 | 50 | 272 | 335.85 | 434.66 | null | 0 | null | null | null |
native-dispatch-001-modal-l4-p1 | modal | L4 | NVIDIA L4 | 7aa089db4b574bf77f4eca092af33188 | data/raw/native-dispatch-001-modal-l4-p1-20260811T205922Z-p117.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 1 | cuda_device_graph | placement | 50 | 50 | 50 | 50 | 16,239.5 | 20,088.05 | 21,616.84 | complete_parent_and_child_execution_environment | 50 | 10,240 | 13,875.2 | 16,404.48 |
native-dispatch-001-modal-l4-p1 | modal | L4 | NVIDIA L4 | 7aa089db4b574bf77f4eca092af33188 | data/raw/native-dispatch-001-modal-l4-p1-20260811T205922Z-p117.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 1 | cuda_host_graph | placement | 50 | 50 | 50 | 50 | 10,298 | 15,068.3 | 16,500.17 | transition_kernels | 50 | 5,120 | 9,216 | 11,317.12 |
native-dispatch-001-modal-l4-p1 | modal | L4 | NVIDIA L4 | 7aa089db4b574bf77f4eca092af33188 | data/raw/native-dispatch-001-modal-l4-p1-20260811T205922Z-p117.csv | 2ec4c6246c6c6287f443c749ebe6385ee9690a9a4eed00a9119cb96aad024395 | 32 | 1 | cuda_host_launch | placement | 50 | 50 | 50 | 50 | 11,208 | 16,387.95 | 20,466.75 | transition_kernels | 50 | 6,144 | 9,779.2 | 14,378.24 |
End of preview.
Ready Cohorts: processed evidence
This is the public processed-evidence mirror for Ready Cohorts: Bounding GPU Opportunity and Avoiding Host Round Trips in LLM-Agent Control by Josef Chen, Independent Researcher.
- Paper: https://arxiv.org/abs/2608.12123
- Hugging Face Paper page: https://huggingface.co/papers/2608.12123
- Code: https://github.com/josefchen/ready-cohorts
- Frozen v1 source and release bundle: https://github.com/josefchen/ready-cohorts/releases/tag/ready-cohorts-arxiv-v1
- Interactive results explorer: https://huggingface.co/spaces/josefchen/ready-cohorts
- This mirror: https://huggingface.co/datasets/josefchen/ready-cohorts
- Input trace dataset at the pinned conversion commit: https://huggingface.co/datasets/Exgentic/agent-llm-traces/tree/f7c94012d0bfbf66fe4d6ed627699508bbb555ff
The input trace bytes are pinned to dataset revision
70036b93a04e61b0ea2706a68b962f4f26774587 and Parquet conversion revision
f7c94012d0bfbf66fe4d6ed627699508bbb555ff. This mirror does not redistribute
prompt text, tool arguments, tool results, API keys, or provider receipts.
Contents
source/: source provenance, session summaries, and derived span features.trace/: the 540 cell-seed evaluations and summary surface for exact equal-relative-deadline packing.resident/: four-placement summaries and paired mechanism contrasts for the device-resident decision study.native/: five-placement summaries and paired contrasts for the fixed nested device-launch negative control.paper/: the release PDF, paper-data manifest, and claim-evidence map.figures/: the three data figures used by the paper.MANIFEST.json: source and public SHA-256 values for every mirrored file.
The resident and native timing rows are technical repetitions. Placement is the outer performance unit. The trace replay is conditional on one 851-session panel, one arrival model, and nine generated swarms. These files do not support hardware-population significance or an end-to-end agent-speedup claim.
Headline cells
At the prospectively frozen trace condition with 100,000 target active
sessions, K=256, and a 50 ms launch deadline, the fixed-window share is
30.19%, the exact offline share is 43.00%, and the local upper bound is 45.85%.
In the separate mechanism study, the device-resident path is faster than the
matched host-round-trip path in all 36 named-placement cells; within-placement
row-median ratios range from 1.19x to 2.39x.
Citation
@misc{chen2026readycohortsboundinggpu,
title = {Ready Cohorts: Bounding GPU Opportunity and Avoiding Host Round Trips in LLM-Agent Control},
author = {Josef Liyanjun Chen},
year = {2026},
eprint = {2608.12123},
archivePrefix = {arXiv},
primaryClass = {cs.DC},
doi = {10.48550/arXiv.2608.12123},
url = {https://arxiv.org/abs/2608.12123}
}
No additional license is asserted by this mirror; upstream source data remain subject to their source repository terms.
- Downloads last month
- 33