AI Infrastructure — Selected Work
Collection
Selected public work: honest inference demos, portfolio, and upcoming reproducible benchmarks for AI Infrastructure Engineer roles. • 4 items • Updated
Dataset Viewer
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
request_id: string
phase: string
ordinal: int64
concurrency: int64
batch_size: int64
prompt_ids: list<item: string>
child 0, item: string
prompt_categories: list<item: string>
child 0, item: string
ok: bool
stats_included: bool
error_type: null
error_message: null
t_start_us: int64
t_finish_us: int64
ttft_ms: double
e2e_ms: double
itl_ms: list<item: double>
child 0, item: double
token_offset_ms: list<item: double>
child 0, item: double
output_tokens: int64
backend: string
model_id: string
output_text: string
meta: struct<run_id: string, run_name: string, created_at: timestamp[s], created_at_is_frozen: bool, packa (... 217 chars omitted)
child 0, run_id: string
child 1, run_name: string
child 2, created_at: timestamp[s]
child 3, created_at_is_frozen: bool
child 4, package_version: string
child 5, synthetic: bool
child 6, seed: int64
child 7, clock: string
child 8, clock_origin: string
child 9, warmup_excluded_from_stats: bool
child 10, percentile_method: string
child 11, closed_loop: bool
child 12, backend: string
child 13, model_id: string
child 14, revision: null
config: struct<schema_version: int64, run_name: string, clock: string, backend: struct<type: string, model_i (... 543 chars omitted)
child 0, schema_version: int64
child 1, run_name: string
child 2, clock: string
child 3, backend: struct<type: string, model_id: string, revision: null, injected_errors: list<item: struct<phase: str (... 98 chars omitted)
child 0, typ
...
tl_ms: string, e2e_ms: string, requests_per_sec: string, output_tokens_per_ (... 148 chars omitted)
child 0, ttft_ms: string
child 1, itl_ms: string
child 2, e2e_ms: string
child 3, requests_per_sec: string
child 4, output_tokens_per_sec: string
child 5, mean_decode_tokens_per_sec: string
child 6, measurement_window_s: string
child 7, output_tokens: string
child 8, error_rate: string
child 9, percentile_method: string
hardware: struct<measurable: bool, peak_allocated_mb: null, instantaneous_used_mb: null, source: null, note: s (... 6 chars omitted)
child 0, measurable: bool
child 1, peak_allocated_mb: null
child 2, instantaneous_used_mb: null
child 3, source: null
child 4, note: string
schema_version: int64
warmup_errors: list<item: null>
child 0, item: null
errors: list<item: struct<request_id: string, phase: string, ordinal: int64, concurrency: int64, batch_size: (... 77 chars omitted)
child 0, item: struct<request_id: string, phase: string, ordinal: int64, concurrency: int64, batch_size: int64, pro (... 65 chars omitted)
child 0, request_id: string
child 1, phase: string
child 2, ordinal: int64
child 3, concurrency: int64
child 4, batch_size: int64
child 5, prompt_ids: list<item: string>
child 0, item: string
child 6, error_type: string
child 7, message: string
notice: string
limitations: list<item: string>
child 0, item: string
warnings: list<item: string>
child 0, item: string
to
{'notice': Value('string'), 'schema_version': Value('int64'), 'meta': {'run_id': Value('string'), 'run_name': Value('string'), 'created_at': Value('timestamp[s]'), 'created_at_is_frozen': Value('bool'), 'package_version': Value('string'), 'synthetic': Value('bool'), 'seed': Value('int64'), 'clock': Value('string'), 'clock_origin': Value('string'), 'warmup_excluded_from_stats': Value('bool'), 'percentile_method': Value('string'), 'closed_loop': Value('bool'), 'backend': Value('string'), 'model_id': Value('string'), 'revision': Value('null')}, 'config': {'schema_version': Value('int64'), 'run_name': Value('string'), 'clock': Value('string'), 'backend': {'type': Value('string'), 'model_id': Value('string'), 'revision': Value('null'), 'injected_errors': List({'phase': Value('string'), 'concurrency': Value('int64'), 'batch_size': Value('int64'), 'ordinal': Value('int64'), 'error_type': Value('string'), 'message': Value('string')})}, 'workload': {'prompts': Value('string'), 'concurrency': List(Value('int64')), 'batch_sizes': List(Value('int64')), 'warmup': Value('int64'), 'samples': Value('int64'), 'max_tokens': Value('int64'), 'temperature': Value('float64'), 'seed': Value('int64'), 'timeout_s': Value('float64')}, 'output': {'directory': Value('string'), 'overwrite': Value('bool'), 'run_id': Value('string'), 'freeze_created_at': Value('timestamp[s]'), 'include_output_text': Value('bool'), 'also_copy_summary_to': List(Value('string'))}}, 'prompts': {'path': Value('string'), 'count'
...
4'), 'measurement_window_s': Value('float64'), 'requests_per_sec': Value('float64'), 'output_tokens_per_sec': Value('float64'), 'mean_decode_tokens_per_sec': Value('float64'), 'n_decode_samples': Value('int64'), 'ttft_ms': {'n': Value('int64'), 'mean': Value('float64'), 'p50': Value('float64'), 'p95': Value('float64'), 'min': Value('float64'), 'max': Value('float64')}, 'itl_ms': {'n': Value('int64'), 'mean': Value('float64'), 'p50': Value('float64'), 'p95': Value('float64')}, 'e2e_ms': {'n': Value('int64'), 'mean': Value('float64'), 'p50': Value('float64'), 'p95': Value('float64'), 'p99': Value('float64'), 'min': Value('float64'), 'max': Value('float64')}, 'e2e_histogram': {'bin_edges_ms': List(Value('float64')), 'counts': List(Value('int64'))}, 'e2e_cdf': List({'e2e_ms': Value('float64'), 'fraction': Value('float64')})}), 'errors': List({'request_id': Value('string'), 'phase': Value('string'), 'ordinal': Value('int64'), 'concurrency': Value('int64'), 'batch_size': Value('int64'), 'prompt_ids': List(Value('string')), 'error_type': Value('string'), 'message': Value('string')}), 'warmup_errors': List(Value('null')), 'definitions': {'ttft_ms': Value('string'), 'itl_ms': Value('string'), 'e2e_ms': Value('string'), 'requests_per_sec': Value('string'), 'output_tokens_per_sec': Value('string'), 'mean_decode_tokens_per_sec': Value('string'), 'measurement_window_s': Value('string'), 'output_tokens': Value('string'), 'error_rate': Value('string'), 'percentile_method': 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
request_id: string
phase: string
ordinal: int64
concurrency: int64
batch_size: int64
prompt_ids: list<item: string>
child 0, item: string
prompt_categories: list<item: string>
child 0, item: string
ok: bool
stats_included: bool
error_type: null
error_message: null
t_start_us: int64
t_finish_us: int64
ttft_ms: double
e2e_ms: double
itl_ms: list<item: double>
child 0, item: double
token_offset_ms: list<item: double>
child 0, item: double
output_tokens: int64
backend: string
model_id: string
output_text: string
meta: struct<run_id: string, run_name: string, created_at: timestamp[s], created_at_is_frozen: bool, packa (... 217 chars omitted)
child 0, run_id: string
child 1, run_name: string
child 2, created_at: timestamp[s]
child 3, created_at_is_frozen: bool
child 4, package_version: string
child 5, synthetic: bool
child 6, seed: int64
child 7, clock: string
child 8, clock_origin: string
child 9, warmup_excluded_from_stats: bool
child 10, percentile_method: string
child 11, closed_loop: bool
child 12, backend: string
child 13, model_id: string
child 14, revision: null
config: struct<schema_version: int64, run_name: string, clock: string, backend: struct<type: string, model_i (... 543 chars omitted)
child 0, schema_version: int64
child 1, run_name: string
child 2, clock: string
child 3, backend: struct<type: string, model_id: string, revision: null, injected_errors: list<item: struct<phase: str (... 98 chars omitted)
child 0, typ
...
tl_ms: string, e2e_ms: string, requests_per_sec: string, output_tokens_per_ (... 148 chars omitted)
child 0, ttft_ms: string
child 1, itl_ms: string
child 2, e2e_ms: string
child 3, requests_per_sec: string
child 4, output_tokens_per_sec: string
child 5, mean_decode_tokens_per_sec: string
child 6, measurement_window_s: string
child 7, output_tokens: string
child 8, error_rate: string
child 9, percentile_method: string
hardware: struct<measurable: bool, peak_allocated_mb: null, instantaneous_used_mb: null, source: null, note: s (... 6 chars omitted)
child 0, measurable: bool
child 1, peak_allocated_mb: null
child 2, instantaneous_used_mb: null
child 3, source: null
child 4, note: string
schema_version: int64
warmup_errors: list<item: null>
child 0, item: null
errors: list<item: struct<request_id: string, phase: string, ordinal: int64, concurrency: int64, batch_size: (... 77 chars omitted)
child 0, item: struct<request_id: string, phase: string, ordinal: int64, concurrency: int64, batch_size: int64, pro (... 65 chars omitted)
child 0, request_id: string
child 1, phase: string
child 2, ordinal: int64
child 3, concurrency: int64
child 4, batch_size: int64
child 5, prompt_ids: list<item: string>
child 0, item: string
child 6, error_type: string
child 7, message: string
notice: string
limitations: list<item: string>
child 0, item: string
warnings: list<item: string>
child 0, item: string
to
{'notice': Value('string'), 'schema_version': Value('int64'), 'meta': {'run_id': Value('string'), 'run_name': Value('string'), 'created_at': Value('timestamp[s]'), 'created_at_is_frozen': Value('bool'), 'package_version': Value('string'), 'synthetic': Value('bool'), 'seed': Value('int64'), 'clock': Value('string'), 'clock_origin': Value('string'), 'warmup_excluded_from_stats': Value('bool'), 'percentile_method': Value('string'), 'closed_loop': Value('bool'), 'backend': Value('string'), 'model_id': Value('string'), 'revision': Value('null')}, 'config': {'schema_version': Value('int64'), 'run_name': Value('string'), 'clock': Value('string'), 'backend': {'type': Value('string'), 'model_id': Value('string'), 'revision': Value('null'), 'injected_errors': List({'phase': Value('string'), 'concurrency': Value('int64'), 'batch_size': Value('int64'), 'ordinal': Value('int64'), 'error_type': Value('string'), 'message': Value('string')})}, 'workload': {'prompts': Value('string'), 'concurrency': List(Value('int64')), 'batch_sizes': List(Value('int64')), 'warmup': Value('int64'), 'samples': Value('int64'), 'max_tokens': Value('int64'), 'temperature': Value('float64'), 'seed': Value('int64'), 'timeout_s': Value('float64')}, 'output': {'directory': Value('string'), 'overwrite': Value('bool'), 'run_id': Value('string'), 'freeze_created_at': Value('timestamp[s]'), 'include_output_text': Value('bool'), 'also_copy_summary_to': List(Value('string'))}}, 'prompts': {'path': Value('string'), 'count'
...
4'), 'measurement_window_s': Value('float64'), 'requests_per_sec': Value('float64'), 'output_tokens_per_sec': Value('float64'), 'mean_decode_tokens_per_sec': Value('float64'), 'n_decode_samples': Value('int64'), 'ttft_ms': {'n': Value('int64'), 'mean': Value('float64'), 'p50': Value('float64'), 'p95': Value('float64'), 'min': Value('float64'), 'max': Value('float64')}, 'itl_ms': {'n': Value('int64'), 'mean': Value('float64'), 'p50': Value('float64'), 'p95': Value('float64')}, 'e2e_ms': {'n': Value('int64'), 'mean': Value('float64'), 'p50': Value('float64'), 'p95': Value('float64'), 'p99': Value('float64'), 'min': Value('float64'), 'max': Value('float64')}, 'e2e_histogram': {'bin_edges_ms': List(Value('float64')), 'counts': List(Value('int64'))}, 'e2e_cdf': List({'e2e_ms': Value('float64'), 'fraction': Value('float64')})}), 'errors': List({'request_id': Value('string'), 'phase': Value('string'), 'ordinal': Value('int64'), 'concurrency': Value('int64'), 'batch_size': Value('int64'), 'prompt_ids': List(Value('string')), 'error_type': Value('string'), 'message': Value('string')}), 'warmup_errors': List(Value('null')), 'definitions': {'ttft_ms': Value('string'), 'itl_ms': Value('string'), 'e2e_ms': Value('string'), 'requests_per_sec': Value('string'), 'output_tokens_per_sec': Value('string'), 'mean_decode_tokens_per_sec': Value('string'), 'measurement_window_s': Value('string'), 'output_tokens': Value('string'), 'error_rate': Value('string'), 'percentile_method': Value('string')}}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Inference Benchmark Results
Raw and summary exports from the LLM Inference Benchmark Lab (benchlab).
Dashboard: https://huggingface.co/spaces/Ki-Asif/inference-benchmark-lab Source: https://github.com/asif-reh/llm-inference-benchmark
Real run (default)
data/local-cpu-smollm2-135m/ — REAL measurement, meta.synthetic=false.
- Model:
HuggingFaceTB/SmolLM2-135M-Instruct - Serving: local
transformersreference HTTP/SSE server on CPU - Workload: 20 fixed prompts; concurrency 1, 2, 4; 8 measured samples per level; 2 warmup excluded
- Files:
summary.json,summary.csv,requests.jsonl,config.yaml - Limits: N=8 per level, serialized generation lock, no GPU/VRAM numbers
Headline client-side results (2026-11-05 UTC):
| Concurrency | TTFT p50 | End-to-end p50 | Errors |
|---|---|---|---|
| 1 | 111.2 ms | 1141 ms | 0/8 |
| 2 | 1096.7 ms | 1626 ms | 0/8 |
| 4 | 3465.4 ms | 4515 ms | 0/8 |
Synthetic example
data/example-synthetic.json — SYNTHETIC EXAMPLE — NOT A REAL BENCHMARK.
Produced by the mock backend's discrete-event clock so the dashboard and schema are reproducible without a GPU or API token. Do not cite these numbers as hardware or model-server performance.
Schema
Summary JSON schema_version=1: meta, config, levels (concurrency sweeps), errors, methodology notes.
- Downloads last month
- 68