| { |
| "engine": { |
| "runtime_env": "LLADA_RUNTIME_ROOT", |
| "engine_revision": "f2230a5c2fcca72440d6c3fea4b50441b7d7516c", |
| "image": "docker.io/lmsysorg/sglang@sha256:065d403274324cc9dec4f830ca62540e6ac586746e1aed268c98a2b347d5853f", |
| "algorithm": "SDARPrefixKV", |
| "tensor_parallel_size": 1, |
| "request_concurrency": 1, |
| "mem_fraction_static": 0.6, |
| "attention_backend": "flashinfer", |
| "disable_radix_cache": true, |
| "disable_cuda_graph": false, |
| "cuda_graph_batch_sizes": [ |
| 1, |
| 2 |
| ], |
| "context_length": null, |
| "context_policy": "No explicit override in verified multi_gen engine; supplied model/runtime default" |
| }, |
| "sampling": { |
| "temperature": 0.0, |
| "temperature_min": 0.0, |
| "temperature_max": 0.0, |
| "top_k": 0, |
| "top_p": 1.0 |
| }, |
| "decoder_common": { |
| "confidence_temperature": 1.0, |
| "confidence_temperature_mode": "fixed", |
| "accept_rule": "prob_threshold", |
| "edit_threshold": 0.5, |
| "block_size": 32, |
| "denoising_steps": 32, |
| "max_steps_per_block": 32, |
| "force_accept_leftmost": true, |
| "allow_accepted_change": true, |
| "post_accept_refine_steps": 16, |
| "active_input_mode": "exact", |
| "adjacent_entropy_contraction": true, |
| "edit_shield_eos": false, |
| "edit_shield_eos_window": 0, |
| "deterministic_randtok": true, |
| "randtok_seed": 12345, |
| "logits_shift": false, |
| "noise_mode": "mask", |
| "random_token_high": 156895, |
| "self_conditioning_post_norm": false, |
| "self_conditioning_residual_scale": 1.0, |
| "causal_prefill": true, |
| "block_causal_prefill": true, |
| "prompt_tail_in_first_block": true, |
| "commit_mode": "fused_next" |
| }, |
| "role": { |
| "decoder": { |
| "prob_threshold": 0.85 |
| }, |
| "benchmark_decoder": {}, |
| "max_new_tokens": { |
| "gsm8k": 4096, |
| "math500": 4096, |
| "humaneval": 4096, |
| "mbpp": 4096, |
| "aime": 16384, |
| "gpqa": 16384, |
| "mmlu_pro": 4096, |
| "lcb": 4096, |
| "ifeval": 2048 |
| } |
| } |
| } |
|
|
