# file: /home/runner/work/FlashSpec/FlashSpec/benchmarks/harness.py
# hypothesis_version: 6.170.11

[0.1, 0.5, 1.0, 100.0, 1000.0, 512, 1024, 2048, '"', '#', "'", ',', ', ', '- ', '--short', '. ', ':', 'False', 'HEAD', 'Null', 'True', '_compiled', '_from_model_config', '_inc', 'acceptance_rate_mean', 'acceptance_rate_std', 'alpaca', 'apply_chat_template', 'arm_report', 'arms', 'ascii', 'assemble_result', 'assistant_model', 'auto', 'avx2', 'bandit', 'bench-drafter', 'bf16', 'bfloat16', 'by_kind', 'c', 'call_shapes', 'chat_template', 'cold_start', 'compile', 'compile_config', 'compile_fixed_shapes', 'compiled', 'config', 'content', 'context_tokens', 'counter', 'cpu', 'cuda', 'cuda_graphs', 'cuda_sync_debug_mode', 'dataset', 'datasets', 'decision', 'decoding', 'definition', 'detect_hardware', 'device', 'do_sample', 'draft_ms_mean', 'drafter', 'drafter_decode_ms', 'drafter_entries', 'drafter_forward_ms', 'drafter_forwards', 'drafter_hidden', 'drafter_layers', 'drafter_module', 'drafter_strategy', 'drafters', 'dtype', 'eager', 'eagle', 'effective', 'end_to_end', 'eos_token_id', 'error', 'fallback_gamma', 'fallback_reason', 'false', 'first_call_ms', 'first_divergence', 'first_in_process', 'fixed_shapes', 'flashspec_compile', 'flashspec_compiled', 'flashspec_thompson', 'flashspec_ucb', 'float16', 'float32', 'forward_cost', 'forward_counts', 'fp16', 'fp32', 'gamma', 'generate', 'generation_config', 'git', 'git_commit', 'gpu', 'gpu_count', 'gpu_memory_gb', 'graph_call_shapes', 'greedy', 'greedy_token_match', 'half', 'hardware', 'hash', 'hf_assisted', 'hf_greedy', 'hf_greedy_compiled', 'hf_sample', 'hf_sample_compiled', 'host_syncs', 'id', 'inclusive', 'initializer_range', 'input_ids', 'label', 'lookup', 'machine', 'max_new_tokens', 'mean', 'measured', 'measurement_kind', 'median', 'medusa', 'method', 'method_order', 'methods', 'metrics', 'min_new_tokens', 'mode', 'model', 'model_name', 'n_generations', 'n_measurement_steps', 'n_prompts', 'n_samples', 'n_steps', 'n_tokens', 'n_warmup', 'n_warmup_steps', 'nonfinite_logits', 'notes', 'null', 'num_assistant_tokens', 'outputs', 'p50_latency_ms', 'p95_latency_ms', 'p99_latency_ms', 'pad_token_id', 'pad_widths', 'past_key_values', 'per_step', 'per_target_forward', 'percentile', 'phase6', 'platform', 'prompt_format', 'prompt_id', 'prompt_ids', 'prompt_length', 'prompt_tokens', 'prompts', 'prompts_matched', 'python', 'raw', 'reason', 'reference', 'replaced_with', 'reps', 'requested', 'resolve_device', 'resolved', 'rev-parse', 'role', 'runner', 'sample', 'scope', 'seed', 'sha256', 'shipped_drafter', 'shipped_target', 'source', 'speedup_vs_ar_mean', 'speedup_vs_ar_std', 'status', 'std', 'step_cost', 'step_profile', 'stop_at_eos', 'summarise_runs', 'synthetic', 'synthetic_prompts', 'target', 'target_decode_ms', 'target_entries', 'target_forwards', 'target_hidden', 'target_layers', 'target_module', 'target_verify_ms', 'temperature', 'temperatures', 'text', 'timestamp', 'token_ids', 'tokenizer', 'tokens_emitted', 'top_k', 'top_p', 'torch', 'torch.', 'total', 'toy', 'toy_llama', 'transformers_version', 'true', 'ucb1', 'unknown', 'unmeasured', 'use_cache', 'user', 'utf-8', 'vanilla_ar', 'verify_ms_mean', 'vocab_size', 'warmup', '~']