Offline Engine API#
SGLang provides a direct inference engine without the need for an HTTP server, especially for use cases where additional HTTP server adds unnecessary complexity or overhead. Here are two general use cases:
Offline Batch Inference
Custom Server on Top of the Engine
This document focuses on the offline batch inference, demonstrating four different inference modes:
Non-streaming synchronous generation
Streaming synchronous generation
Non-streaming asynchronous generation
Streaming asynchronous generation
Additionally, you can easily build a custom server on top of the SGLang offline engine. A detailed example working in a python script can be found in custom_server.
Nest Asyncio#
Note that if you want to use Offline Engine in ipython or some other nested loop code, you need to add the following code:
import nest_asyncio
nest_asyncio.apply()
Advanced Usage#
The engine supports vlm inference as well as extracting hidden states.
Please see the examples for further use cases.
Offline Batch Inference#
SGLang offline engine supports batch inference with efficient scheduling.
[1]:
# launch the offline engine
import asyncio
import sglang as sgl
import sglang.test.doc_patch
from sglang.utils import async_stream_and_merge, stream_and_merge
llm = sgl.Engine(model_path="qwen/qwen2.5-0.5b-instruct")
[2026-01-24 15:09:32] INFO utils.py:148: Note: detected 112 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
[2026-01-24 15:09:32] INFO utils.py:151: Note: NumExpr detected 112 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
[2026-01-24 15:09:32] INFO utils.py:164: NumExpr defaulting to 16 threads.
<frozen importlib._bootstrap_external>:1184: FutureWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.
<frozen importlib._bootstrap_external>:1184: FutureWarning: The cuda.nvrtc module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.nvrtc module instead.
[2026-01-24 15:09:35] INFO server_args.py:1771: Attention backend not specified. Use fa3 backend by default.
[2026-01-24 15:09:35] INFO server_args.py:2679: Set soft_watchdog_timeout since in CI
[2026-01-24 15:09:35] INFO engine.py:154: server_args=ServerArgs(model_path='qwen/qwen2.5-0.5b-instruct', tokenizer_path='qwen/qwen2.5-0.5b-instruct', tokenizer_mode='auto', tokenizer_worker_num=1, skip_tokenizer_init=False, load_format='auto', model_loader_extra_config='{}', trust_remote_code=False, context_length=None, is_embedding=False, enable_multimodal=None, revision=None, model_impl='auto', host='127.0.0.1', port=30000, fastapi_root_path='', grpc_mode=False, skip_server_warmup=False, warmups=None, nccl_port=None, checkpoint_engine_wait_weights_before_ready=False, dtype='auto', quantization=None, quantization_param_path=None, kv_cache_dtype='auto', enable_fp32_lm_head=False, modelopt_quant=None, modelopt_checkpoint_restore_path=None, modelopt_checkpoint_save_path=None, modelopt_export_path=None, quantize_and_serve=False, rl_quant_profile=None, mem_fraction_static=0.835, max_running_requests=128, max_queued_requests=None, max_total_tokens=20480, chunked_prefill_size=8192, enable_dynamic_chunking=False, max_prefill_tokens=16384, prefill_max_requests=None, schedule_policy='fcfs', enable_priority_scheduling=False, abort_on_priority_when_disabled=False, schedule_low_priority_values_first=False, priority_scheduling_preemption_threshold=10, schedule_conservativeness=1.0, page_size=1, swa_full_tokens_ratio=0.8, disable_hybrid_swa_memory=False, radix_eviction_policy='lru', enable_prefill_delayer=False, prefill_delayer_max_delay_passes=30, prefill_delayer_token_usage_low_watermark=None, prefill_delayer_forward_passes_buckets=None, prefill_delayer_wait_seconds_buckets=None, device='cuda', tp_size=1, pp_size=1, pp_max_micro_batch_size=None, pp_async_batch_depth=0, stream_interval=1, stream_output=False, random_seed=218839002, constrained_json_whitespace_pattern=None, constrained_json_disable_any_whitespace=False, watchdog_timeout=300, soft_watchdog_timeout=300, dist_timeout=None, download_dir=None, model_checksum=None, base_gpu_id=0, gpu_id_step=1, sleep_on_idle=False, custom_sigquit_handler=None, log_level='error', log_level_http=None, log_requests=False, log_requests_level=2, log_requests_format='text', log_requests_target=None, uvicorn_access_log_exclude_prefixes=[], crash_dump_folder=None, show_time_cost=False, enable_metrics=False, enable_metrics_for_all_schedulers=False, tokenizer_metrics_custom_labels_header='x-custom-labels', tokenizer_metrics_allowed_custom_labels=None, bucket_time_to_first_token=None, bucket_inter_token_latency=None, bucket_e2e_request_latency=None, collect_tokens_histogram=False, prompt_tokens_buckets=None, generation_tokens_buckets=None, gc_warning_threshold_secs=0.0, decode_log_interval=40, enable_request_time_stats_logging=False, kv_events_config=None, enable_trace=False, otlp_traces_endpoint='localhost:4317', export_metrics_to_file=False, export_metrics_to_file_dir=None, api_key=None, admin_api_key=None, served_model_name='qwen/qwen2.5-0.5b-instruct', weight_version='default', chat_template=None, hf_chat_template_name=None, completion_template=None, file_storage_path='sglang_storage', enable_cache_report=False, reasoning_parser=None, tool_call_parser=None, tool_server=None, sampling_defaults='model', dp_size=1, load_balance_method='round_robin', dist_init_addr=None, nnodes=1, node_rank=0, json_model_override_args='{}', preferred_sampling_params=None, enable_lora=None, enable_lora_overlap_loading=None, max_lora_rank=None, lora_target_modules=None, lora_paths=None, max_loaded_loras=None, max_loras_per_batch=8, lora_eviction_policy='lru', lora_backend='csgmv', max_lora_chunk_size=16, attention_backend='fa3', decode_attention_backend=None, prefill_attention_backend=None, sampling_backend='flashinfer', grammar_backend='xgrammar', mm_attention_backend=None, fp8_gemm_runner_backend='auto', fp4_gemm_runner_backend='auto', nsa_prefill_backend=None, nsa_decode_backend=None, disable_flashinfer_autotune=False, speculative_algorithm=None, speculative_draft_model_path=None, speculative_draft_model_revision=None, speculative_draft_load_format=None, speculative_num_steps=None, speculative_eagle_topk=None, speculative_num_draft_tokens=None, speculative_accept_threshold_single=1.0, speculative_accept_threshold_acc=1.0, speculative_token_map=None, speculative_attention_mode='prefill', speculative_draft_attention_backend=None, speculative_moe_runner_backend='auto', speculative_moe_a2a_backend=None, speculative_draft_model_quantization=None, speculative_ngram_min_match_window_size=1, speculative_ngram_max_match_window_size=12, speculative_ngram_min_bfs_breadth=1, speculative_ngram_max_bfs_breadth=10, speculative_ngram_match_type='BFS', speculative_ngram_branch_length=18, speculative_ngram_capacity=10000000, enable_multi_layer_eagle=False, ep_size=1, moe_a2a_backend='none', moe_runner_backend='auto', flashinfer_mxfp4_moe_precision='default', enable_flashinfer_allreduce_fusion=False, deepep_mode='auto', ep_num_redundant_experts=0, ep_dispatch_algorithm=None, init_expert_location='trivial', enable_eplb=False, eplb_algorithm='auto', eplb_rebalance_num_iterations=1000, eplb_rebalance_layers_per_chunk=None, eplb_min_rebalancing_utilization_threshold=1.0, expert_distribution_recorder_mode=None, expert_distribution_recorder_buffer_size=1000, enable_expert_distribution_metrics=False, deepep_config=None, moe_dense_tp_size=None, elastic_ep_backend=None, mooncake_ib_device=None, max_mamba_cache_size=None, mamba_ssm_dtype='float32', mamba_full_memory_ratio=0.9, mamba_scheduler_strategy='no_buffer', mamba_track_interval=256, enable_hierarchical_cache=False, hicache_ratio=2.0, hicache_size=0, hicache_write_policy='write_through', hicache_io_backend='kernel', hicache_mem_layout='layer_first', disable_hicache_numa_detect=False, hicache_storage_backend=None, hicache_storage_prefetch_policy='best_effort', hicache_storage_backend_extra_config=None, hierarchical_sparse_attention_extra_config=None, enable_lmcache=False, kt_weight_path=None, kt_method=None, kt_cpuinfer=None, kt_threadpool_count=None, kt_num_gpu_experts=None, kt_max_deferred_experts_per_token=None, dllm_algorithm=None, dllm_algorithm_config=None, enable_double_sparsity=False, ds_channel_config_path=None, ds_heavy_channel_num=32, ds_heavy_token_num=256, ds_heavy_channel_type='qk', ds_sparse_decode_threshold=4096, cpu_offload_gb=0, offload_group_size=-1, offload_num_in_group=1, offload_prefetch_step=1, offload_mode='cpu', multi_item_scoring_delimiter=None, disable_radix_cache=False, cuda_graph_max_bs=4, cuda_graph_bs=[1, 2, 4, 8, 12, 16, 24, 32, 40, 48, 56, 64, 72, 80, 88, 96, 104, 112, 120, 128, 136, 144, 152, 160, 168, 176, 184, 192, 200, 208, 216, 224, 232, 240, 248, 256], disable_cuda_graph=False, disable_cuda_graph_padding=False, enable_profile_cuda_graph=False, enable_cudagraph_gc=False, enable_layerwise_nvtx_marker=False, enable_nccl_nvls=False, enable_symm_mem=False, disable_flashinfer_cutlass_moe_fp4_allgather=False, enable_tokenizer_batch_encode=False, disable_tokenizer_batch_decode=False, disable_outlines_disk_cache=False, disable_custom_all_reduce=False, enable_mscclpp=False, enable_torch_symm_mem=False, disable_overlap_schedule=False, enable_mixed_chunk=False, enable_dp_attention=False, enable_dp_lm_head=False, enable_two_batch_overlap=False, enable_single_batch_overlap=False, tbo_token_distribution_threshold=0.48, enable_torch_compile=False, enable_piecewise_cuda_graph=False, enable_torch_compile_debug_mode=False, torch_compile_max_bs=32, piecewise_cuda_graph_max_tokens=8192, piecewise_cuda_graph_tokens=[4, 8, 12, 16, 20, 24, 28, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224, 240, 256, 288, 320, 352, 384, 416, 448, 480, 512, 640, 704, 768, 832, 896, 960, 1024, 1280, 1536, 1792, 2048, 2304, 2560, 2816, 3072, 3328, 3584, 3840, 4096, 4608, 5120, 5632, 6144, 6656, 7168, 7680, 8192], piecewise_cuda_graph_compiler='eager', torchao_config='', enable_nan_detection=False, enable_p2p_check=False, triton_attention_reduce_in_fp32=False, triton_attention_num_kv_splits=8, triton_attention_split_tile_size=None, num_continuous_decode_steps=1, delete_ckpt_after_loading=False, enable_memory_saver=False, enable_weights_cpu_backup=False, enable_draft_weights_cpu_backup=False, allow_auto_truncate=False, enable_custom_logit_processor=False, flashinfer_mla_disable_ragged=False, disable_shared_experts_fusion=False, disable_chunked_prefix_cache=False, disable_fast_image_processor=False, keep_mm_feature_on_device=False, enable_return_hidden_states=False, enable_return_routed_experts=False, scheduler_recv_interval=1, numa_node=None, enable_deterministic_inference=False, rl_on_policy_target=None, enable_attn_tp_input_scattered=False, enable_nsa_prefill_context_parallel=False, nsa_prefill_cp_mode='in-seq-split', enable_fused_qk_norm_rope=False, enable_precise_embedding_interpolation=False, enable_dynamic_batch_tokenizer=False, dynamic_batch_tokenizer_batch_size=32, dynamic_batch_tokenizer_batch_timeout=0.002, debug_tensor_dump_output_folder=None, debug_tensor_dump_layers=None, debug_tensor_dump_input_file=None, debug_tensor_dump_inject=False, disaggregation_mode='null', disaggregation_transfer_backend='mooncake', disaggregation_bootstrap_port=8998, disaggregation_decode_tp=None, disaggregation_decode_dp=None, disaggregation_prefill_pp=1, disaggregation_ib_device=None, disaggregation_decode_enable_offload_kvcache=False, disaggregation_decode_enable_fake_auto=False, num_reserved_decode_tokens=512, disaggregation_decode_polling_interval=1, encoder_only=False, language_only=False, encoder_transfer_backend='zmq_to_scheduler', encoder_urls=[], custom_weight_loader=[], weight_loader_disable_mmap=False, remote_instance_weight_loader_seed_instance_ip=None, remote_instance_weight_loader_seed_instance_service_port=None, remote_instance_weight_loader_send_weights_group_ports=None, remote_instance_weight_loader_backend='nccl', remote_instance_weight_loader_start_seed_via_transfer_engine=False, enable_pdmux=False, pdmux_config_path=None, sm_group_num=8, mm_max_concurrent_calls=32, mm_per_request_timeout=10.0, enable_broadcast_mm_inputs_process=False, enable_prefix_mm_cache=False, mm_enable_dp_encoder=False, mm_process_config={}, limit_mm_data_per_request=None, decrypted_config_file=None, decrypted_draft_config_file=None, forward_hooks=None)
<frozen importlib._bootstrap_external>:1184: FutureWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.
<frozen importlib._bootstrap_external>:1184: FutureWarning: The cuda.nvrtc module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.nvrtc module instead.
<frozen importlib._bootstrap_external>:1184: FutureWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.
<frozen importlib._bootstrap_external>:1184: FutureWarning: The cuda.nvrtc module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.nvrtc module instead.
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
Loading safetensors checkpoint shards: 0% Completed | 0/1 [00:00<?, ?it/s]
Loading safetensors checkpoint shards: 100% Completed | 1/1 [00:00<00:00, 5.97it/s]
Loading safetensors checkpoint shards: 100% Completed | 1/1 [00:00<00:00, 5.96it/s]
Capturing batches (bs=1 avail_mem=76.74 GB): 100%|██████████| 20/20 [00:01<00:00, 19.45it/s]
Non-streaming Synchronous Generation#
[2]:
prompts = [
"Hello, my name is",
"The president of the United States is",
"The capital of France is",
"The future of AI is",
]
sampling_params = {"temperature": 0.8, "top_p": 0.95}
outputs = llm.generate(prompts, sampling_params)
for prompt, output in zip(prompts, outputs):
print("===============================")
print(f"Prompt: {prompt}\nGenerated text: {output['text']}")
===============================
Prompt: Hello, my name is
Generated text: John. I am currently living in the United States. My job is as a software developer. I have been working with web applications for a long time. I have a passion for learning new things and always keep learning new skills. I have been learning JavaScript and Node.js for a while and I was wondering if you could give me some advice on how to start learning and improve my skills. Thank you!
Sure, I'd be happy to help you with that! Learning a new programming language or skill can be a rewarding experience, and it's great that you're taking the time to learn. Here are some tips to help you get started
===============================
Prompt: The president of the United States is
Generated text: a man. The president of the United States is elected by the citizens. The president of the United States serves a term of two years. The president of the United States is the most important person in the country. The president of the United States is the leader of the country. The president of the United States is the commander of the army. He is the commander of the navy. He is the commander of the air force. He is the president of the supreme court. He is the chief executive of the government. He is the supreme leader of the country. The president of the United States is responsible for making important decisions. He is
===============================
Prompt: The capital of France is
Generated text: Paris. Which of the following statements about the cultural features of Paris is correct?
A. The residences of important people in Paris are mostly painted in red.
B. Parisian Frenchmen are more reserved and considerate in their behavior.
C. There are four seasons in Paris.
D. Parisian Frenchmen generally have a warm and kind personality.
Answer:
B
What are the main types of floral decorations?
A. Live flowers, artificial flowers, floral wreaths, and floral tassels
B. Live flowers, artificial flowers, floral tassels, and floral cages
C. Live flowers, floral tassels
===============================
Prompt: The future of AI is
Generated text: not clear, but the main tools of future AI will be algorithms and machine learning.
Which of the following statements is correct?
A: The core of the future of AI is to be used.
B: The core of the future of AI is algorithms and machine learning.
C: The future of AI will not only have algorithms but also machine learning.
D: The future of AI will only have algorithms. Let's analyze each option step by step to determine which statement is correct.
A: The core of the future of AI is to be used.
- This statement suggests that the primary goal of future AI is to be used, but it
Streaming Synchronous Generation#
[3]:
prompts = [
"Write a short, neutral self-introduction for a fictional character. Hello, my name is",
"Provide a concise factual statement about France’s capital city. The capital of France is",
"Explain possible future trends in artificial intelligence. The future of AI is",
]
sampling_params = {
"temperature": 0.2,
"top_p": 0.9,
}
print("\n=== Testing synchronous streaming generation with overlap removal ===\n")
for prompt in prompts:
print(f"Prompt: {prompt}")
merged_output = stream_and_merge(llm, prompt, sampling_params)
print("Generated text:", merged_output)
print()
=== Testing synchronous streaming generation with overlap removal ===
Prompt: Write a short, neutral self-introduction for a fictional character. Hello, my name is
Generated text: [Name], and I'm a [job title] at [company name]. I'm excited to meet you and learn more about you. What can you tell me about yourself? I'm a [job title] at [company name], and I'm passionate about [job title] and I love [job title] because [reason for passion]. I'm always looking for new challenges and opportunities to grow and learn. What's your favorite hobby or activity? I love [hobby or activity], and I'm always looking for new ways to explore and discover new things. What's your favorite book or movie? I love [book/movie
Prompt: Provide a concise factual statement about France’s capital city. The capital of France is
Generated text: Paris, the city known for its iconic landmarks such as the Eiffel Tower, Notre-Dame Cathedral, and the Louvre Museum. It is also home to the French Parliament, the French National Library, and the French National Museum of Modern Art. Paris is a bustling metropolis with a rich cultural heritage and is a popular tourist destination. The city is known for its cuisine, fashion, and art, and is a major center for business and finance in Europe. Paris is a city of contrasts, with its modern architecture and historical landmarks blending seamlessly. Its status as the capital of France is a testament to its importance as a cultural and
Prompt: Explain possible future trends in artificial intelligence. The future of AI is
Generated text: likely to be characterized by a number of trends that are expected to shape the way we interact with technology and the world around us. Here are some possible future trends in AI:
1. Increased automation and robotics: As AI technology continues to advance, we can expect to see more automation and robotics in various industries. This could lead to increased efficiency, reduced costs, and improved quality of life for many people.
2. AI in healthcare: AI is already being used in healthcare to improve patient outcomes and reduce costs. As AI technology continues to advance, we can expect to see even more sophisticated applications in healthcare, such as personalized medicine, disease prediction
Non-streaming Asynchronous Generation#
[4]:
prompts = [
"Write a short, neutral self-introduction for a fictional character. Hello, my name is",
"Provide a concise factual statement about France’s capital city. The capital of France is",
"Explain possible future trends in artificial intelligence. The future of AI is",
]
sampling_params = {"temperature": 0.8, "top_p": 0.95}
print("\n=== Testing asynchronous batch generation ===")
async def main():
outputs = await llm.async_generate(prompts, sampling_params)
for prompt, output in zip(prompts, outputs):
print(f"\nPrompt: {prompt}")
print(f"Generated text: {output['text']}")
asyncio.run(main())
=== Testing asynchronous batch generation ===
Prompt: Write a short, neutral self-introduction for a fictional character. Hello, my name is
Generated text: [insert name], and I am a [insert occupation or field] expert who is passionate about [insert topic or area of interest].
As a [insert occupation or field] expert who is passionate about [insert topic or area of interest], I believe that my expertise lies in [insert area of expertise]. My [insert area of expertise] has been honed through years of [insert specific skills or experience], and I am eager to share this knowledge with others.
In my free time, I enjoy [insert hobbies or interests], such as [insert hobbies or interests]. I am also a [insert occupation or field] enthusiast who is always
Prompt: Provide a concise factual statement about France’s capital city. The capital of France is
Generated text: Paris, the largest city in the country and its political, cultural, and economic center. The city has a rich history dating back to the medieval times, and it is known for its iconic landmarks such as the Eiffel Tower, Notre-Dame Cathedral, and Montmartre. It is also a major tourist destination, with Paris being recognized as a world-renowned city for its art, architecture, and food. The city is home to numerous cultural institutions, including the Louvre Museum and the Musée d'Orsay, as well as the Palace of Versailles, which is known for its opulent furnishings and lavish gardens.
Prompt: Explain possible future trends in artificial intelligence. The future of AI is
Generated text: promising, with many possibilities for growth and innovation. Here are some potential trends in AI that could shape the field in the coming years:
1. Increased transparency and accountability: With more transparency and accountability, AI systems will become more trustworthy and predictable, which could lead to a more stable and reliable AI market.
2. Enhanced personalization: AI will continue to gain personalization capabilities, allowing for more tailored and contextually relevant services and recommendations. This could lead to more efficient and effective use of resources.
3. Autonomous decision-making: Autonomous AI systems will become more autonomous and capable, capable of making decisions based on data-driven algorithms rather than human
Streaming Asynchronous Generation#
[5]:
prompts = [
"Write a short, neutral self-introduction for a fictional character. Hello, my name is",
"Provide a concise factual statement about France’s capital city. The capital of France is",
"Explain possible future trends in artificial intelligence. The future of AI is",
]
sampling_params = {"temperature": 0.8, "top_p": 0.95}
print("\n=== Testing asynchronous streaming generation (no repeats) ===")
async def main():
for prompt in prompts:
print(f"\nPrompt: {prompt}")
print("Generated text: ", end="", flush=True)
# Replace direct calls to async_generate with our custom overlap-aware version
async for cleaned_chunk in async_stream_and_merge(llm, prompt, sampling_params):
print(cleaned_chunk, end="", flush=True)
print() # New line after each prompt
asyncio.run(main())
=== Testing asynchronous streaming generation (no repeats) ===
Prompt: Write a short, neutral self-introduction for a fictional character. Hello, my name is
Generated text: [Name] and I'm a [Age] year old [Name] who has always been interested in [occupation or hobby]. I'm an [extroverted or introverted] person who thrives in [work environment or personal setting]. I enjoy [what do you love to do?], and I enjoy [what else do you love to do?]. I'm [able to do [number of things] on my own, and [able to do [number of things] with others]. I've been [career goal or project], and I'm [what I plan to do in the next [number of months or years
Prompt: Provide a concise factual statement about France’s capital city. The capital of France is
Generated text: Paris.
Paris is the capital city of France. It is the largest city in France and the third largest in Europe, with an estimated population of over 2.2 million people. The city is famous for its rich cultural heritage, fashion, gastronomy, and nightlife. Paris is known for its iconic landmarks such as the Eiffel Tower, Notre-Dame Cathedral, and the Louvre Museum. It is also home to many world-renowned institutions and landmarks, including the Sorbonne University, the Louvre Museum, and the Opera House. Paris is a major transportation hub, with two major airports and a wide range of public
Prompt: Explain possible future trends in artificial intelligence. The future of AI is
Generated text: likely to be shaped by a number of key trends, including:
1. Self-learning: AI systems will become more adept at learning from their interactions with the world, rather than being programmed by humans. This means that AI will be able to improve on its own, without the need for human intervention.
2. Personalization: As AI systems become more sophisticated, they will be able to adapt to individual users and provide tailored experiences. This means that AI will be able to offer more personalized recommendations, targeted advertising, and other forms of personalization.
3. Transparency: The way AI systems are developed and used will become more transparent, as developers
[6]:
llm.shutdown()