Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
path: string
sha256: string
first_line: int64
last_line: int64
selection_sha256: string
outputs: struct<train: struct<path: string, sha256: string>, test: struct<path: string, sha256: string>>
  child 0, train: struct<path: string, sha256: string>
      child 0, path: string
      child 1, sha256: string
  child 1, test: struct<path: string, sha256: string>
      child 0, path: string
      child 1, sha256: string
role: string
counts: struct<train:prefill: int64, test:prefill: int64, train:decode: int64, test:decode: int64, train:mix (... 29 chars omitted)
  child 0, train:prefill: int64
  child 1, test:prefill: int64
  child 2, train:decode: int64
  child 3, test:decode: int64
  child 4, train:mixed: int64
  child 5, test:mixed: int64
split: string
excluded: struct<outside_selected_ranges: int64, idle_in_selected_ranges: int64>
  child 0, outside_selected_ranges: int64
  child 1, idle_in_selected_ranges: int64
capture_report_sha256: string
to
{'selection_sha256': Value('string'), 'capture_report_sha256': Value('string'), 'role': Value('string'), 'split': Value('string'), 'counts': {'train:prefill': Value('int64'), 'test:prefill': Value('int64'), 'train:decode': Value('int64'), 'test:decode': Value('int64'), 'train:mixed': Value('int64'), 'test:mixed': Value('int64')}, 'excluded': {'outside_selected_ranges': Value('int64'), 'idle_in_selected_ranges': Value('int64')}, 'outputs': {'train': {'path': Value('string'), 'sha256': Value('string')}, 'test': {'path': Value('string'), 'sha256': Value('string')}}}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              path: string
              sha256: string
              first_line: int64
              last_line: int64
              selection_sha256: string
              outputs: struct<train: struct<path: string, sha256: string>, test: struct<path: string, sha256: string>>
                child 0, train: struct<path: string, sha256: string>
                    child 0, path: string
                    child 1, sha256: string
                child 1, test: struct<path: string, sha256: string>
                    child 0, path: string
                    child 1, sha256: string
              role: string
              counts: struct<train:prefill: int64, test:prefill: int64, train:decode: int64, test:decode: int64, train:mix (... 29 chars omitted)
                child 0, train:prefill: int64
                child 1, test:prefill: int64
                child 2, train:decode: int64
                child 3, test:decode: int64
                child 4, train:mixed: int64
                child 5, test:mixed: int64
              split: string
              excluded: struct<outside_selected_ranges: int64, idle_in_selected_ranges: int64>
                child 0, outside_selected_ranges: int64
                child 1, idle_in_selected_ranges: int64
              capture_report_sha256: string
              to
              {'selection_sha256': Value('string'), 'capture_report_sha256': Value('string'), 'role': Value('string'), 'split': Value('string'), 'counts': {'train:prefill': Value('int64'), 'test:prefill': Value('int64'), 'train:decode': Value('int64'), 'test:decode': Value('int64'), 'train:mixed': Value('int64'), 'test:mixed': Value('int64')}, 'excluded': {'outside_selected_ranges': Value('int64'), 'idle_in_selected_ranges': Value('int64')}, 'outputs': {'train': {'path': Value('string'), 'sha256': Value('string')}, 'test': {'path': Value('string'), 'sha256': Value('string')}}}
              because column names don't match

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.

AISim dataset

Measured data used by AISim to predict LLM serving performance, in one repository with one layout and one discovery index. Pin a full commit SHA, not a branch name, when you use it.

Three categories:

  • op-latency: per-operation latency tables per system, backend and version.
  • fpm: whole-forward-pass (FPM) execution profiles, including native step captures.
  • e2e: end-to-end serving measurements (observations, metrics, per-request rows and time series) of aisim-collected workloads, such as the AgentX Weka workload on Qwen3.5-397B-A17B (B300, vLLM 0.31.0, TP8).

Layout

catalog/dataset.json          dataset index: categories, their roots and category-index artifacts (with SHA-256)
catalog/e2e/index.json        E2E category index: one entry per configuration (1)
catalog/fpm/index.json        FPM category index: one entry per configuration (62)
catalog/op-latency.json       op-latency category index: one entry per version directory (884)
catalog/<layer>/<hardware>.parquet   coverage catalog shards (rebuilt by aisim_core.data.catalog.build_catalog)
catalog/systems.parquet       system id -> spec hash
catalog/op-latency-reuse.parquet     declared op-table reuse with donors resolved
catalog/schemas/              JSON Schema snapshots of every registered aisim-* schema
data/fpm/<org>--<model>/<hardware>/<parallelism>/...   whole-forward-pass (FPM) execution profiles
data/e2e/<org>--<model>/<hardware>/<backend>/<version>/<parallelism>/cfg-<id>/   observations, metrics, requests, series
data/fpm/history/...          superseded FPM snapshots (envelopes exist; not in the default index)
data/op-latency/<system>/<family>/<backend>/<version>/ per-operation latency tables
data/op-latency/source_map.parquet    source-to-published file map with hashes
audit/, docs/, DSV41_*.md     provenance notes carried over from the previous FPM repository

Every configuration directory has an aisim-manifest.json (aisim-artifact-manifest/v1) that lists each file with role, SHA-256, size and Parquet row count. check_dataset re-hashes every artifact the indexes reference.

Notes for consumers:

  • FPM paths: files moved from data/<rest> to data/fpm/<rest> byte-for-byte. Embedded strings in moved metadata still say data/<model>/...; treat configuration_path as a logical key and apply the data/fpm/ prefix.
  • Op-latency Parquet tables are zstd-compressed and carry the aisim-op-latency/v1 schema ID; YAML sources are embedded verbatim in the converted Parquet metadata.

Schemas

Registered schema IDs (aisim-dataset-index/v1, aisim-category-index/v1, aisim-artifact-manifest/v1, aisim-catalog/v1, aisim-op-*/v1, ...) have JSON Schema snapshots in catalog/schemas/.

Fetching

Download only what a selection needs, verified against the hashes in the manifests:

aisim data fetch --category fpm --model deepseek-ai/DeepSeek-V4-Pro --hardware b300_sxm \
    --repo ai-dynamo/aisim-data --revision <full-commit-sha> --dry-run
aisim data fetch --category op-latency --hardware h200_sxm --framework trtllm \
    --repo ai-dynamo/aisim-data --revision <full-commit-sha>

To make AISim use the data in normal commands, pin it by environment variable:

export AISIM_DATASET=ai-dynamo/aisim-data@<full-commit-sha>
export AISIM_DATA_CATEGORIES=fpm,op-latency

A local checkout laid out as above can be used without network via AISIM_DATA_LOCAL=<dir>. Check a copy with aisim_core.data.check_dataset(<dir>).

License

Pending. The license for this dataset is not decided; it depends on the OSRB review tracked in AISIM-2311 and AISIM-2312. No license is granted or implied by this dataset, and no LICENSE file is included. Do not redistribute until it is set.

The replayed AgentX workload corpus (semianalysisai/cc-traces-weka-062126, revision 23f152f6) is Apache-2.0; this is recorded in the e2e manifest (workload.dataset.license).

Downloads last month
176