Inference-Perf CLI Flags
July 1, 2026 · View on GitHub
These command line flags are automatically generated from the internal Config schema. You can override any configuration directly from the CLI without using a yaml configuration file.
| Flag | Type | Description |
|---|---|---|
--api.type | Enum (completion, chat) | Matches api.type in config |
--api.streaming | boolean | Matches api.streaming in config |
--api.headers | JSON | Matches api.headers in config |
--api.slo_unit | str | Matches api.slo_unit in config |
--api.slo_tpot_header | str | Matches api.slo_tpot_header in config |
--api.slo_ttft_header | str | Matches api.slo_ttft_header in config |
--api.response_format.type | Enum (json_schema, json_object) | Matches api.response_format.type in config |
--api.response_format.name | str | Matches api.response_format.name in config |
--api.response_format.json_schema | JSON | Matches api.response_format.json_schema in config |
--api.session_id_header_key | str | Matches api.session_id_header_key in config |
--data.type | Enum (mock, shareGPT, synthetic, random, shared_prefix, cnn_dailymail, infinity_instruct, billsum_conversations, otel_trace_replay, weka_trace_replay, conversation_replay, visionarena) | Matches data.type in config |
--data.path | str | Matches data.path in config |
--data.corpus_file_path | str | Path to a text file to use as the prompt tokenization corpus instead of the default hardcoded sonnet |
--data.input_distribution.min | int | Matches data.input_distribution.min in config |
--data.input_distribution.max | int | Matches data.input_distribution.max in config |
--data.input_distribution.mean | float | Matches data.input_distribution.mean in config |
--data.input_distribution.std_dev | float | Matches data.input_distribution.std_dev in config |
--data.input_distribution.total_count | int | Matches data.input_distribution.total_count in config |
--data.input_distribution.type | Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.input_distribution.type in config |
--data.input_distribution.variance | float | Matches data.input_distribution.variance in config |
--data.input_distribution.skew | float | Matches data.input_distribution.skew in config |
--data.output_distribution.min | int | Matches data.output_distribution.min in config |
--data.output_distribution.max | int | Matches data.output_distribution.max in config |
--data.output_distribution.mean | float | Matches data.output_distribution.mean in config |
--data.output_distribution.std_dev | float | Matches data.output_distribution.std_dev in config |
--data.output_distribution.total_count | int | Matches data.output_distribution.total_count in config |
--data.output_distribution.type | Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.output_distribution.type in config |
--data.output_distribution.variance | float | Matches data.output_distribution.variance in config |
--data.output_distribution.skew | float | Matches data.output_distribution.skew in config |
--data.shared_prefix.num_groups | int | Matches data.shared_prefix.num_groups in config |
--data.shared_prefix.num_prompts_per_group | int | Matches data.shared_prefix.num_prompts_per_group in config |
--data.shared_prefix.system_prompt_len | string | Matches data.shared_prefix.system_prompt_len in config |
--data.shared_prefix.question_len | string | Matches data.shared_prefix.question_len in config |
--data.shared_prefix.output_len | string | Matches data.shared_prefix.output_len in config |
--data.shared_prefix.seed | int | Matches data.shared_prefix.seed in config |
--data.shared_prefix.question_distribution.min | int | Matches data.shared_prefix.question_distribution.min in config |
--data.shared_prefix.question_distribution.max | int | Matches data.shared_prefix.question_distribution.max in config |
--data.shared_prefix.question_distribution.mean | float | Matches data.shared_prefix.question_distribution.mean in config |
--data.shared_prefix.question_distribution.std_dev | float | Matches data.shared_prefix.question_distribution.std_dev in config |
--data.shared_prefix.question_distribution.total_count | int | Matches data.shared_prefix.question_distribution.total_count in config |
--data.shared_prefix.question_distribution.type | Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.shared_prefix.question_distribution.type in config |
--data.shared_prefix.question_distribution.variance | float | Matches data.shared_prefix.question_distribution.variance in config |
--data.shared_prefix.question_distribution.skew | float | Matches data.shared_prefix.question_distribution.skew in config |
--data.shared_prefix.output_distribution.min | int | Matches data.shared_prefix.output_distribution.min in config |
--data.shared_prefix.output_distribution.max | int | Matches data.shared_prefix.output_distribution.max in config |
--data.shared_prefix.output_distribution.mean | float | Matches data.shared_prefix.output_distribution.mean in config |
--data.shared_prefix.output_distribution.std_dev | float | Matches data.shared_prefix.output_distribution.std_dev in config |
--data.shared_prefix.output_distribution.total_count | int | Matches data.shared_prefix.output_distribution.total_count in config |
--data.shared_prefix.output_distribution.type | Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.shared_prefix.output_distribution.type in config |
--data.shared_prefix.output_distribution.variance | float | Matches data.shared_prefix.output_distribution.variance in config |
--data.shared_prefix.output_distribution.skew | float | Matches data.shared_prefix.output_distribution.skew in config |
--data.shared_prefix.enable_multi_turn_chat | boolean | Matches data.shared_prefix.enable_multi_turn_chat in config |
--data.shared_prefix.multimodal.image.count.min | int | Matches data.shared_prefix.multimodal.image.count.min in config |
--data.shared_prefix.multimodal.image.count.max | int | Matches data.shared_prefix.multimodal.image.count.max in config |
--data.shared_prefix.multimodal.image.count.mean | float | Matches data.shared_prefix.multimodal.image.count.mean in config |
--data.shared_prefix.multimodal.image.count.std_dev | float | Matches data.shared_prefix.multimodal.image.count.std_dev in config |
--data.shared_prefix.multimodal.image.count.total_count | int | Matches data.shared_prefix.multimodal.image.count.total_count in config |
--data.shared_prefix.multimodal.image.count.type | Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.shared_prefix.multimodal.image.count.type in config |
--data.shared_prefix.multimodal.image.count.variance | float | Matches data.shared_prefix.multimodal.image.count.variance in config |
--data.shared_prefix.multimodal.image.count.skew | float | Matches data.shared_prefix.multimodal.image.count.skew in config |
--data.shared_prefix.multimodal.image.insertion_point | string | Placement of media within the text prompt. Float in range [0.0, 1.0] (0=start, 1=end), or a Distribution to sample from. |
--data.shared_prefix.multimodal.image.resolutions | JSON | Resolution or list of weighted resolutions for generated images. |
--data.shared_prefix.multimodal.image.representation | Enum (png, jpeg, webp) | Wire encoding for emitted image bytes: png (default, lossless) or jpeg (lossy, smaller payload). Some VLMs prefer one or the other; consult the model's spec sheet. |
--data.shared_prefix.multimodal.video.count.min | int | Matches data.shared_prefix.multimodal.video.count.min in config |
--data.shared_prefix.multimodal.video.count.max | int | Matches data.shared_prefix.multimodal.video.count.max in config |
--data.shared_prefix.multimodal.video.count.mean | float | Matches data.shared_prefix.multimodal.video.count.mean in config |
--data.shared_prefix.multimodal.video.count.std_dev | float | Matches data.shared_prefix.multimodal.video.count.std_dev in config |
--data.shared_prefix.multimodal.video.count.total_count | int | Matches data.shared_prefix.multimodal.video.count.total_count in config |
--data.shared_prefix.multimodal.video.count.type | Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.shared_prefix.multimodal.video.count.type in config |
--data.shared_prefix.multimodal.video.count.variance | float | Matches data.shared_prefix.multimodal.video.count.variance in config |
--data.shared_prefix.multimodal.video.count.skew | float | Matches data.shared_prefix.multimodal.video.count.skew in config |
--data.shared_prefix.multimodal.video.insertion_point | string | Placement of media within the text prompt. Float in range [0.0, 1.0] (0=start, 1=end), or a Distribution to sample from. |
--data.shared_prefix.multimodal.video.profiles | JSON | Video profile or list of weighted video profiles for generated videos. |
--data.shared_prefix.multimodal.video.representation | Enum (mp4, png_frames, jpeg_frames) | Wire-format strategy. mp4 sends one video_url block carrying an MP4 blob (measures full pipeline including server-side decode). png_frames and jpeg_frames send frames × image_url blocks at one insertion point in the named encoding (no decode dependency, useful for prefix-cache benchmarks and servers that don't accept video_url). |
--data.shared_prefix.multimodal.audio.count.min | int | Matches data.shared_prefix.multimodal.audio.count.min in config |
--data.shared_prefix.multimodal.audio.count.max | int | Matches data.shared_prefix.multimodal.audio.count.max in config |
--data.shared_prefix.multimodal.audio.count.mean | float | Matches data.shared_prefix.multimodal.audio.count.mean in config |
--data.shared_prefix.multimodal.audio.count.std_dev | float | Matches data.shared_prefix.multimodal.audio.count.std_dev in config |
--data.shared_prefix.multimodal.audio.count.total_count | int | Matches data.shared_prefix.multimodal.audio.count.total_count in config |
--data.shared_prefix.multimodal.audio.count.type | Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.shared_prefix.multimodal.audio.count.type in config |
--data.shared_prefix.multimodal.audio.count.variance | float | Matches data.shared_prefix.multimodal.audio.count.variance in config |
--data.shared_prefix.multimodal.audio.count.skew | float | Matches data.shared_prefix.multimodal.audio.count.skew in config |
--data.shared_prefix.multimodal.audio.insertion_point | string | Placement of media within the text prompt. Float in range [0.0, 1.0] (0=start, 1=end), or a Distribution to sample from. |
--data.shared_prefix.multimodal.audio.durations | JSON | Duration or list of weighted durations for generated audio clips. |
--data.multimodal.image.count.min | int | Matches data.multimodal.image.count.min in config |
--data.multimodal.image.count.max | int | Matches data.multimodal.image.count.max in config |
--data.multimodal.image.count.mean | float | Matches data.multimodal.image.count.mean in config |
--data.multimodal.image.count.std_dev | float | Matches data.multimodal.image.count.std_dev in config |
--data.multimodal.image.count.total_count | int | Matches data.multimodal.image.count.total_count in config |
--data.multimodal.image.count.type | Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.multimodal.image.count.type in config |
--data.multimodal.image.count.variance | float | Matches data.multimodal.image.count.variance in config |
--data.multimodal.image.count.skew | float | Matches data.multimodal.image.count.skew in config |
--data.multimodal.image.insertion_point | string | Placement of media within the text prompt. Float in range [0.0, 1.0] (0=start, 1=end), or a Distribution to sample from. |
--data.multimodal.image.resolutions | JSON | Resolution or list of weighted resolutions for generated images. |
--data.multimodal.image.representation | Enum (png, jpeg, webp) | Wire encoding for emitted image bytes: png (default, lossless) or jpeg (lossy, smaller payload). Some VLMs prefer one or the other; consult the model's spec sheet. |
--data.multimodal.video.count.min | int | Matches data.multimodal.video.count.min in config |
--data.multimodal.video.count.max | int | Matches data.multimodal.video.count.max in config |
--data.multimodal.video.count.mean | float | Matches data.multimodal.video.count.mean in config |
--data.multimodal.video.count.std_dev | float | Matches data.multimodal.video.count.std_dev in config |
--data.multimodal.video.count.total_count | int | Matches data.multimodal.video.count.total_count in config |
--data.multimodal.video.count.type | Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.multimodal.video.count.type in config |
--data.multimodal.video.count.variance | float | Matches data.multimodal.video.count.variance in config |
--data.multimodal.video.count.skew | float | Matches data.multimodal.video.count.skew in config |
--data.multimodal.video.insertion_point | string | Placement of media within the text prompt. Float in range [0.0, 1.0] (0=start, 1=end), or a Distribution to sample from. |
--data.multimodal.video.profiles | JSON | Video profile or list of weighted video profiles for generated videos. |
--data.multimodal.video.representation | Enum (mp4, png_frames, jpeg_frames) | Wire-format strategy. mp4 sends one video_url block carrying an MP4 blob (measures full pipeline including server-side decode). png_frames and jpeg_frames send frames × image_url blocks at one insertion point in the named encoding (no decode dependency, useful for prefix-cache benchmarks and servers that don't accept video_url). |
--data.multimodal.audio.count.min | int | Matches data.multimodal.audio.count.min in config |
--data.multimodal.audio.count.max | int | Matches data.multimodal.audio.count.max in config |
--data.multimodal.audio.count.mean | float | Matches data.multimodal.audio.count.mean in config |
--data.multimodal.audio.count.std_dev | float | Matches data.multimodal.audio.count.std_dev in config |
--data.multimodal.audio.count.total_count | int | Matches data.multimodal.audio.count.total_count in config |
--data.multimodal.audio.count.type | Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.multimodal.audio.count.type in config |
--data.multimodal.audio.count.variance | float | Matches data.multimodal.audio.count.variance in config |
--data.multimodal.audio.count.skew | float | Matches data.multimodal.audio.count.skew in config |
--data.multimodal.audio.insertion_point | string | Placement of media within the text prompt. Float in range [0.0, 1.0] (0=start, 1=end), or a Distribution to sample from. |
--data.multimodal.audio.durations | JSON | Duration or list of weighted durations for generated audio clips. |
--data.trace.file | str | Matches data.trace.file in config |
--data.trace.format | Enum (AzurePublicDataset) | Matches data.trace.format in config |
--data.otel_trace_replay.use_static_model | boolean | Use a single static model for all requests |
--data.otel_trace_replay.static_model_name | str | Static model name (required if use_static_model=True) |
--data.otel_trace_replay.model_mapping | JSON | Map recorded model names to target models |
--data.otel_trace_replay.default_max_tokens | int | Default max_tokens if not specified in trace |
--data.otel_trace_replay.override_tool_call_max_tokens | boolean | Override tool call max_tokens to 4096 instead of using trace recorded length |
--data.otel_trace_replay.inject_random_session_id | boolean | Inject random string into unique segments to invalidate KV-cache between sessions |
--data.otel_trace_replay.duplicate_sessions_target | int | Target number of sessions to reach by duplicating existing sessions. If None, no duplication occurs. |
--data.otel_trace_replay.max_wait_ms | int | Maximum inter-event wait time in milliseconds. Caps the delay between predecessor completion and event dispatch to avoid reproducing unusually long tool/agent execution times from the original trace. |
--data.otel_trace_replay.include_errors | boolean | Include spans with error status |
--data.otel_trace_replay.skip_invalid_files | boolean | Skip invalid trace files instead of failing |
--data.otel_trace_replay.trace_directory | str | Directory containing OTel JSON trace files |
--data.otel_trace_replay.trace_files | JSON | List of paths to specific OTel JSON trace files |
--data.otel_trace_replay.hf_dataset_path | JSON | HuggingFace dataset path. Can be: |
- String: 'username/dataset-name'
- Dict: {'path': 'username/dataset-name', 'revision': 'main', 'split': 'train'}
Any extra keys in the dict are passed as kwargs to datasets.load_dataset(). |
|
--data.otel_trace_replay.filter| str | Lambda expression to filter trace records. Applied uniformly to all data sources. Example: "lambda x: x['benchmark'] == 'gsm8k'" or "lambda x: 'spans' in x and len(x['spans']) > 5" Security: Filter expressions use eval() and should only contain trusted input. | |--data.otel_trace_replay.attribute_to_header_map| JSON | Map OTel span attributes to HTTP headers | |--data.otel_trace_replay.attribute_to_label_map| JSON | Map OTel span attributes to metrics reporting labels | |--data.otel_trace_replay.bad_tool_call_handling| Enum (none, use_recorded) | How to handle tool_calls whose function.arguments is not valid JSON. none (default): no mitigation, bytes propagate and vLLM may return HTTP 400 on the next turn. use_recorded: discard the live response and substitute the recorded assistant message at the affected slot; the recorded tool_call_id flows into the recorded role:tool successor unchanged. | |--data.weka_trace_replay.use_static_model| boolean | Use a single static model for all requests | |--data.weka_trace_replay.static_model_name| str | Static model name (required if use_static_model=True) | |--data.weka_trace_replay.model_mapping| JSON | Map recorded model names to target models | |--data.weka_trace_replay.default_max_tokens| int | Default max_tokens if not specified in trace | |--data.weka_trace_replay.override_tool_call_max_tokens| boolean | Override tool call max_tokens to 4096 instead of using trace recorded length | |--data.weka_trace_replay.inject_random_session_id| boolean | Inject random string into unique segments to invalidate KV-cache between sessions | |--data.weka_trace_replay.duplicate_sessions_target| int | Target number of sessions to reach by duplicating existing sessions. If None, no duplication occurs. | |--data.weka_trace_replay.max_wait_ms| int | Maximum inter-event wait time in milliseconds. Caps the delay between predecessor completion and event dispatch to avoid reproducing unusually long tool/agent execution times from the original trace. | |--data.weka_trace_replay.include_errors| boolean | Include spans with error status | |--data.weka_trace_replay.skip_invalid_files| boolean | Skip invalid trace files instead of failing | |--data.weka_trace_replay.trace_directory| str | Directory containing Weka JSON trace files | |--data.weka_trace_replay.trace_files| JSON | List of paths to specific Weka JSON trace files | |--data.weka_trace_replay.hf_dataset_path| JSON | HuggingFace dataset path. Can be: - String: 'username/dataset-name'
- Dict: {'path': 'username/dataset-name', 'revision': 'main', 'split': 'train'}
Any extra keys in the dict are passed as kwargs to datasets.load_dataset(). |
|
--data.weka_trace_replay.trace_idle_gap_cap_seconds| float | Cap idle timing gaps between turns in seconds | |--data.weka_trace_replay.ignore_trace_delays| boolean | Ignore delays/delays from original trace and run back-to-back | |--data.weka_trace_replay.use_think_time_only| boolean | Only use think_time attribute instead of timestamps | |--data.weka_trace_replay.default_block_size| int | Default block size if not specified in trace | |--data.weka_trace_replay.num_dataset_entries| int | Max number of dataset traces to load from HuggingFace | |--data.conversation_replay.seed| int | Random seed for deterministic generation | |--data.conversation_replay.num_conversations| int | Number of conversation blueprints to generate | |--data.conversation_replay.shared_system_prompt_len| int | Fixed shared system prompt length in tokens | |--data.conversation_replay.dynamic_system_prompt_len.min| int | Matches data.conversation_replay.dynamic_system_prompt_len.min in config | |--data.conversation_replay.dynamic_system_prompt_len.max| int | Matches data.conversation_replay.dynamic_system_prompt_len.max in config | |--data.conversation_replay.dynamic_system_prompt_len.mean| float | Matches data.conversation_replay.dynamic_system_prompt_len.mean in config | |--data.conversation_replay.dynamic_system_prompt_len.std_dev| float | Matches data.conversation_replay.dynamic_system_prompt_len.std_dev in config | |--data.conversation_replay.dynamic_system_prompt_len.total_count| int | Matches data.conversation_replay.dynamic_system_prompt_len.total_count in config | |--data.conversation_replay.dynamic_system_prompt_len.type| Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.conversation_replay.dynamic_system_prompt_len.type in config | |--data.conversation_replay.dynamic_system_prompt_len.variance| float | Matches data.conversation_replay.dynamic_system_prompt_len.variance in config | |--data.conversation_replay.dynamic_system_prompt_len.skew| float | Matches data.conversation_replay.dynamic_system_prompt_len.skew in config | |--data.conversation_replay.turns_per_conversation.min| int | Matches data.conversation_replay.turns_per_conversation.min in config | |--data.conversation_replay.turns_per_conversation.max| int | Matches data.conversation_replay.turns_per_conversation.max in config | |--data.conversation_replay.turns_per_conversation.mean| float | Matches data.conversation_replay.turns_per_conversation.mean in config | |--data.conversation_replay.turns_per_conversation.std_dev| float | Matches data.conversation_replay.turns_per_conversation.std_dev in config | |--data.conversation_replay.turns_per_conversation.total_count| int | Matches data.conversation_replay.turns_per_conversation.total_count in config | |--data.conversation_replay.turns_per_conversation.type| Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.conversation_replay.turns_per_conversation.type in config | |--data.conversation_replay.turns_per_conversation.variance| float | Matches data.conversation_replay.turns_per_conversation.variance in config | |--data.conversation_replay.turns_per_conversation.skew| float | Matches data.conversation_replay.turns_per_conversation.skew in config | |--data.conversation_replay.input_tokens_per_turn.min| int | Matches data.conversation_replay.input_tokens_per_turn.min in config | |--data.conversation_replay.input_tokens_per_turn.max| int | Matches data.conversation_replay.input_tokens_per_turn.max in config | |--data.conversation_replay.input_tokens_per_turn.mean| float | Matches data.conversation_replay.input_tokens_per_turn.mean in config | |--data.conversation_replay.input_tokens_per_turn.std_dev| float | Matches data.conversation_replay.input_tokens_per_turn.std_dev in config | |--data.conversation_replay.input_tokens_per_turn.total_count| int | Matches data.conversation_replay.input_tokens_per_turn.total_count in config | |--data.conversation_replay.input_tokens_per_turn.type| Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.conversation_replay.input_tokens_per_turn.type in config | |--data.conversation_replay.input_tokens_per_turn.variance| float | Matches data.conversation_replay.input_tokens_per_turn.variance in config | |--data.conversation_replay.input_tokens_per_turn.skew| float | Matches data.conversation_replay.input_tokens_per_turn.skew in config | |--data.conversation_replay.output_tokens_per_turn.min| int | Matches data.conversation_replay.output_tokens_per_turn.min in config | |--data.conversation_replay.output_tokens_per_turn.max| int | Matches data.conversation_replay.output_tokens_per_turn.max in config | |--data.conversation_replay.output_tokens_per_turn.mean| float | Matches data.conversation_replay.output_tokens_per_turn.mean in config | |--data.conversation_replay.output_tokens_per_turn.std_dev| float | Matches data.conversation_replay.output_tokens_per_turn.std_dev in config | |--data.conversation_replay.output_tokens_per_turn.total_count| int | Matches data.conversation_replay.output_tokens_per_turn.total_count in config | |--data.conversation_replay.output_tokens_per_turn.type| Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.conversation_replay.output_tokens_per_turn.type in config | |--data.conversation_replay.output_tokens_per_turn.variance| float | Matches data.conversation_replay.output_tokens_per_turn.variance in config | |--data.conversation_replay.output_tokens_per_turn.skew| float | Matches data.conversation_replay.output_tokens_per_turn.skew in config | |--data.conversation_replay.tool_call_latency_sec.min| int | Matches data.conversation_replay.tool_call_latency_sec.min in config | |--data.conversation_replay.tool_call_latency_sec.max| int | Matches data.conversation_replay.tool_call_latency_sec.max in config | |--data.conversation_replay.tool_call_latency_sec.mean| float | Matches data.conversation_replay.tool_call_latency_sec.mean in config | |--data.conversation_replay.tool_call_latency_sec.std_dev| float | Matches data.conversation_replay.tool_call_latency_sec.std_dev in config | |--data.conversation_replay.tool_call_latency_sec.total_count| int | Matches data.conversation_replay.tool_call_latency_sec.total_count in config | |--data.conversation_replay.tool_call_latency_sec.type| Enum (normal, skew_normal, lognormal, uniform, poisson, fixed) | Matches data.conversation_replay.tool_call_latency_sec.type in config | |--data.conversation_replay.tool_call_latency_sec.variance| float | Matches data.conversation_replay.tool_call_latency_sec.variance in config | |--data.conversation_replay.tool_call_latency_sec.skew| float | Matches data.conversation_replay.tool_call_latency_sec.skew in config | |--data.conversation_replay.max_model_len| int | Maximum model context length in tokens | |--data.visionarena.hf_dataset_name| str | HuggingFace dataset identifier; override only when mirroring the dataset elsewhere. | |--data.visionarena.hf_split| str | HuggingFace split to stream. | |--data.visionarena.hf_data_files| str | Optionaldata_filesglob forwarded toload_dataset. | |--data.visionarena.num_rows| int | Number of usable rows to stream into the in-memory request pool at startup. Caps memory use; the benchmark cycles through this pool. | |--data.visionarena.max_images_per_request| int | Cap on images attached per request; truncates a row's image list. | |--data.visionarena.insertion_point| string | Placement of the image block(s) within the prompt text. Float in [0.0, 1.0] (0=start, 1=end), or a Distribution to sample per request. | |--load.type| Enum (constant, poisson, trace_replay, concurrent, trace_session_replay) | Matches load.type in config | |--load.interval| float | Matches load.interval in config | |--load.stages| JSON | Matches load.stages in config | |--load.sweep.type| Enum (geometric, linear) | Matches load.sweep.type in config | |--load.sweep.num_requests| int | Matches load.sweep.num_requests in config | |--load.sweep.timeout| float | Matches load.sweep.timeout in config | |--load.sweep.num_stages| int | Matches load.sweep.num_stages in config | |--load.sweep.stage_duration| int | Matches load.sweep.stage_duration in config | |--load.sweep.saturation_percentile| float | Matches load.sweep.saturation_percentile in config | |--load.num_workers| int | Matches load.num_workers in config | |--load.worker_max_concurrency| int | Matches load.worker_max_concurrency in config | |--load.worker_max_tcp_connections| int | Matches load.worker_max_tcp_connections in config | |--load.trace.file| str | Matches load.trace.file in config | |--load.trace.format| Enum (AzurePublicDataset) | Matches load.trace.format in config | |--load.circuit_breakers| JSON | Matches load.circuit_breakers in config | |--load.request_timeout| float | Matches load.request_timeout in config | |--load.lora_traffic_split| JSON | Matches load.lora_traffic_split in config | |--load.base_seed| int | Matches load.base_seed in config | |--metrics.type| Enum (prometheus, default) | Matches metrics.type in config | |--metrics.prometheus.scrape_interval| int | Matches metrics.prometheus.scrape_interval in config | |--metrics.prometheus.url| string | Matches metrics.prometheus.url in config | |--metrics.prometheus.filters| JSON | Matches metrics.prometheus.filters in config | |--metrics.prometheus.google_managed| boolean | Matches metrics.prometheus.google_managed in config | |--report.request_lifecycle.summary| boolean | Matches report.request_lifecycle.summary in config | |--report.request_lifecycle.per_stage| boolean | Matches report.request_lifecycle.per_stage in config | |--report.request_lifecycle.per_request| boolean | Matches report.request_lifecycle.per_request in config | |--report.request_lifecycle.per_adapter| boolean | Matches report.request_lifecycle.per_adapter in config | |--report.request_lifecycle.per_adapter_stage| boolean | Matches report.request_lifecycle.per_adapter_stage in config | |--report.request_lifecycle.percentiles| JSON | Matches report.request_lifecycle.percentiles in config | |--report.request_lifecycle.use_server_output_tokens| boolean | Matches report.request_lifecycle.use_server_output_tokens in config | |--report.prometheus.summary| boolean | Matches report.prometheus.summary in config | |--report.prometheus.per_stage| boolean | Matches report.prometheus.per_stage in config | |--report.session_lifecycle.summary| boolean | Matches report.session_lifecycle.summary in config | |--report.session_lifecycle.per_stage| boolean | Matches report.session_lifecycle.per_stage in config | |--report.session_lifecycle.per_session| boolean | Matches report.session_lifecycle.per_session in config | |--report.goodput.constraints| JSON | Matches report.goodput.constraints in config | |--storage.local_storage.path| str | Matches storage.local_storage.path in config | |--storage.local_storage.report_file_prefix| str | Matches storage.local_storage.report_file_prefix in config | |--storage.google_cloud_storage.path| str | Matches storage.google_cloud_storage.path in config | |--storage.google_cloud_storage.report_file_prefix| str | Matches storage.google_cloud_storage.report_file_prefix in config | |--storage.google_cloud_storage.bucket_name| str | Matches storage.google_cloud_storage.bucket_name in config | |--storage.simple_storage_service.path| str | Matches storage.simple_storage_service.path in config | |--storage.simple_storage_service.report_file_prefix| str | Matches storage.simple_storage_service.report_file_prefix in config | |--storage.simple_storage_service.bucket_name| str | Matches storage.simple_storage_service.bucket_name in config | |--storage.simple_storage_service.endpoint_url| str | Matches storage.simple_storage_service.endpoint_url in config | |--storage.simple_storage_service.region_name| str | Matches storage.simple_storage_service.region_name in config | |--storage.simple_storage_service.addressing_style| string | Matches storage.simple_storage_service.addressing_style in config | |--server.type| Enum (vllm, sglang, tgi, mock) | Matches server.type in config | |--server.model_name| str | Matches server.model_name in config | |--server.base_url| str | Matches server.base_url in config | |--server.ignore_eos| boolean | Matches server.ignore_eos in config | |--server.api_key| str | Matches server.api_key in config | |--server.cert_path| str | Matches server.cert_path in config | |--server.key_path| str | Matches server.key_path in config | |--tokenizer.pretrained_model_name_or_path| str | Matches tokenizer.pretrained_model_name_or_path in config | |--tokenizer.trust_remote_code| boolean | Matches tokenizer.trust_remote_code in config | |--tokenizer.token| str | Matches tokenizer.token in config | |--circuit_breakers| JSON | Matches circuit_breakers in config |