Datasets:

click6067
/

fitllm-fit-census

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
name: string
version: int64
schema_version: int64
generated: timestamp[s]
license: string
license_url: string
rows: int64
devices: int64
measured_rows: int64
canonical: struct<site: string, json: string, csv: string, source: string>
  child 0, site: string
  child 1, json: string
  child 2, csv: string
  child 3, source: string
sha256: struct<census-v1.json: string, census-v1.csv: string>
  child 0, census-v1.json: string
  child 1, census-v1.csv: string
engine: struct<name: string, license: string, npm: string, repo: string>
  child 0, name: string
  child 1, license: string
  child 2, npm: string
  child 3, repo: string
regenerate: string
data: list<item: struct<model: string, params_b: double, device: string, platform: string, memory_gb: int6 (... 465 chars omitted)
  child 0, item: struct<model: string, params_b: double, device: string, platform: string, memory_gb: int64, quant: s (... 453 chars omitted)
      child 0, model: string
      child 1, params_b: double
      child 2, device: string
      child 3, platform: string
      child 4, memory_gb: int64
      child 5, quant: string
      child 6, ctx: int64
      child 7, kv: string
      child 8, verdict: string
      child 9, used_gb: double
      child 10, predicted_total_to_run_gb: double
      child 11, predicted_param_gb: double
      child 12, predicted_resident_weights_gb: double
      child 13, kv_cache_gb: double
      child 14, runtime_dynamic_gb: double
      child 15, reserve_gb: double
      child 16, free_gb: double
      child 17, max_context: int64
      child 18, measured_peak_gb: double
      child 19, measurement_kind: string
      child 20, measured_ctx: int64
      child 21, measurement_match: string
      child 22, measured_unit: string
      child 23, measured_evidence_level: string
      child 24, measured_source: string
assumptions: string
measured_from: string
verdicts: int64
engine_data: string
definitions: struct<predicted_total_to_run_gb: string, predicted_resident_weights_gb: string, measurement_kind: s (... 48 chars omitted)
  child 0, predicted_total_to_run_gb: string
  child 1, predicted_resident_weights_gb: string
  child 2, measurement_kind: string
  child 3, measurement_match: string
  child 4, units: string
to
{'version': Value('int64'), 'schema_version': Value('int64'), 'generated': Value('timestamp[s]'), 'engine_data': Value('string'), 'verdicts': Value('int64'), 'assumptions': Value('string'), 'definitions': {'predicted_total_to_run_gb': Value('string'), 'predicted_resident_weights_gb': Value('string'), 'measurement_kind': Value('string'), 'measurement_match': Value('string'), 'units': Value('string')}, 'regenerate': Value('string'), 'measured_from': Value('string'), 'data': List({'model': Value('string'), 'params_b': Value('float64'), 'device': Value('string'), 'platform': Value('string'), 'memory_gb': Value('int64'), 'quant': Value('string'), 'ctx': Value('int64'), 'kv': Value('string'), 'verdict': Value('string'), 'used_gb': Value('float64'), 'predicted_total_to_run_gb': Value('float64'), 'predicted_param_gb': Value('float64'), 'predicted_resident_weights_gb': Value('float64'), 'kv_cache_gb': Value('float64'), 'runtime_dynamic_gb': Value('float64'), 'reserve_gb': Value('float64'), 'free_gb': Value('float64'), 'max_context': Value('int64'), 'measured_peak_gb': Value('float64'), 'measurement_kind': Value('string'), 'measured_ctx': Value('int64'), 'measurement_match': Value('string'), 'measured_unit': Value('string'), 'measured_evidence_level': Value('string'), 'measured_source': Value('string')})}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, 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 127, 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 478, 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 2818, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2355, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2380, 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 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, 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 343, 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 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
              name: string
              version: int64
              schema_version: int64
              generated: timestamp[s]
              license: string
              license_url: string
              rows: int64
              devices: int64
              measured_rows: int64
              canonical: struct<site: string, json: string, csv: string, source: string>
                child 0, site: string
                child 1, json: string
                child 2, csv: string
                child 3, source: string
              sha256: struct<census-v1.json: string, census-v1.csv: string>
                child 0, census-v1.json: string
                child 1, census-v1.csv: string
              engine: struct<name: string, license: string, npm: string, repo: string>
                child 0, name: string
                child 1, license: string
                child 2, npm: string
                child 3, repo: string
              regenerate: string
              data: list<item: struct<model: string, params_b: double, device: string, platform: string, memory_gb: int6 (... 465 chars omitted)
                child 0, item: struct<model: string, params_b: double, device: string, platform: string, memory_gb: int64, quant: s (... 453 chars omitted)
                    child 0, model: string
                    child 1, params_b: double
                    child 2, device: string
                    child 3, platform: string
                    child 4, memory_gb: int64
                    child 5, quant: string
                    child 6, ctx: int64
                    child 7, kv: string
                    child 8, verdict: string
                    child 9, used_gb: double
                    child 10, predicted_total_to_run_gb: double
                    child 11, predicted_param_gb: double
                    child 12, predicted_resident_weights_gb: double
                    child 13, kv_cache_gb: double
                    child 14, runtime_dynamic_gb: double
                    child 15, reserve_gb: double
                    child 16, free_gb: double
                    child 17, max_context: int64
                    child 18, measured_peak_gb: double
                    child 19, measurement_kind: string
                    child 20, measured_ctx: int64
                    child 21, measurement_match: string
                    child 22, measured_unit: string
                    child 23, measured_evidence_level: string
                    child 24, measured_source: string
              assumptions: string
              measured_from: string
              verdicts: int64
              engine_data: string
              definitions: struct<predicted_total_to_run_gb: string, predicted_resident_weights_gb: string, measurement_kind: s (... 48 chars omitted)
                child 0, predicted_total_to_run_gb: string
                child 1, predicted_resident_weights_gb: string
                child 2, measurement_kind: string
                child 3, measurement_match: string
                child 4, units: string
              to
              {'version': Value('int64'), 'schema_version': Value('int64'), 'generated': Value('timestamp[s]'), 'engine_data': Value('string'), 'verdicts': Value('int64'), 'assumptions': Value('string'), 'definitions': {'predicted_total_to_run_gb': Value('string'), 'predicted_resident_weights_gb': Value('string'), 'measurement_kind': Value('string'), 'measurement_match': Value('string'), 'units': Value('string')}, 'regenerate': Value('string'), 'measured_from': Value('string'), 'data': List({'model': Value('string'), 'params_b': Value('float64'), 'device': Value('string'), 'platform': Value('string'), 'memory_gb': Value('int64'), 'quant': Value('string'), 'ctx': Value('int64'), 'kv': Value('string'), 'verdict': Value('string'), 'used_gb': Value('float64'), 'predicted_total_to_run_gb': Value('float64'), 'predicted_param_gb': Value('float64'), 'predicted_resident_weights_gb': Value('float64'), 'kv_cache_gb': Value('float64'), 'runtime_dynamic_gb': Value('float64'), 'reserve_gb': Value('float64'), 'free_gb': Value('float64'), 'max_context': Value('int64'), 'measured_peak_gb': Value('float64'), 'measurement_kind': Value('string'), 'measured_ctx': Value('int64'), 'measurement_match': Value('string'), 'measured_unit': Value('string'), 'measured_evidence_level': Value('string'), 'measured_source': 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.

FitLLM Fit Census — 6,048 "will it run?" verdicts (CC0)

Every row answers: can this local LLM run on this GPU or Apple Silicon Mac? 19 models × 88 devices (36 GPUs + 52 Mac configs) × per-platform quantization tiers = 6,048 verdicts, each with the full memory breakdown and max context.

Computed by the open-source MIT fitllm-engine from official model config.json values — architecture-aware math (MLA latent KV, sliding-window, hybrid linear attention, MoE, PLE weight residency) that naive 2×layers×heads×head_dim formulas get wrong by up to 17.8×. Hardware numbers are cross-verified against ≥2 independent sources (URLs cited in-engine per value).

manifest.json ships alongside the data: license, row count, sha256 checksums and canonical URLs — cite or pin the dataset without parsing anything else.

Columns

column	meaning
`model`, `params_b`	model name, total params (B)
`device`, `platform`, `memory_gb`	hardware, `gpu`/`mac`, VRAM or unified memory
`quant`, `ctx`, `kv`	weight quant tier, context tokens, KV-cache precision
`verdict`	`yes` / `tight` / `no`
`used_gb`, `free_gb`, `max_context`	memory used/free, max context at this quant
`predicted_total_to_run_gb`, `predicted_resident_weights_gb`, `predicted_param_gb`	typed predictions — total-to-run (what the verdict uses) vs resident weights (compare with `ollama ps`-style readings) vs bare weights
`measured_peak_gb`, `measurement_kind`, `measured_evidence_level`, `measured_source`	community-measured real usage, typed (`idle_resident`/`generation_peak`/…) with evidence grade — only `system_total_peak` compares directly to the total

Reproduce / verify

git clone https://github.com/click6067-ship-it/fitllm-engine && cd fitllm-engine
npm run census        # regenerates this exact dataset (+ manifest.json with sha256)
node vectors/run.mjs  # 15 conformance vectors — byte-exact KV anchors

Live per-combo API (no auth): https://fitllm.run/api/check?model=<name>&gpu=<name> · machine snapshot: https://fitllm.run/data/c-hf.json

License: CC0 1.0 — public domain. Use it, train on it, redistribute it. Attribution appreciated (fitllm.run), never required.

Downloads last month: 89

Total file size:

4.64 MB