sgl-project/sglang
Documented errors, page 24 of 32. Back to sgl-project/sglang
| Code / Message | Type | Severity | Tags |
|---|---|---|---|
| SANA-WM refiner requires batch.latents from stage 1. | validation | error | sana-wm, refiner, missing-latents, pipeline-order, valueerror |
| SANA-WM streaming CFG requires negative prompt embeds. | validation | error | sana-wm, streaming, cfg, negative-prompt, valueerror |
| is not supported on XPU (no XPU kernel implementation). | exception | error | xpu, intel-gpu, quantization, hardware-support, fp8 |
| `sglang.bench_one_batch_server` is deprecated and will be… | console | warning | deprecation, benchmark, one-batch-server, future-warning |
| Unknown speculative algorithm name | validation | error | speculative-decoding, configuration, enum-lookup |
| update_mask length mismatch | validation | error | minimax-h3, update-mask, tensor-parallel, logits |
| Both --speculative-num-draft-tokens and… | validation | error | speculative-decoding, dflash, conflicting-arguments, block-size |
| Cannot forward both `prompt | validation | error | glm-image, prompt-embeds, mutual-exclusion, input-validation, valueerror |
| capacity must be positive, got | validation | error | cuda-graph, kimi-k3, config-validation, vision-tower |
| capped layout has a row exceeding cap= | validation | error | speculative-decoding, validation, capacity-limit |
| caption_valid_mask must have one row per Z-Image caption | validation | error | z-image, mask-validation, batching |
| decode_context_parallel_size | exception | error | parallelism, decode-context-parallel, config-validation, sglang |
| --enable-linear-replayssm-spec is not supported on a PD… | validation | error | sglang, replayssm, pd-disaggregation, speculative-decoding |
| FlashInferKDAKernel has no prefill kernel; keep prefill on… | exception | error | sglang, flashinfer, kda, prefill, not-implemented |
| GGUF and safetensors quantization metadata conflict | validation | error | gguf, quantization, metadata-conflict, load-spec |
| Invalid select_mode: . Choose 'topk' or 'threshold'. | validation | error | validation, value-error, moba, attention, select-mode |
| Kimi image placeholders must map one-to-one to image data… | validation | error | kimi, k25, multimodal, placeholder-count, validation |
| kitchen_int8 group_size must be one of | validation | error | quantization, config-validation, group-size |
| MiniMax H3 base64 decoded size | validation | error | base64, internal-consistency, minimax-h3 |
| Per-layer checkpoint quantization and Nunchaku are mutually… | validation | error | nunchaku, quantization, mutually-exclusive, config-conflict |
| Qwen3-Next shared expert fusion currently supports exactly… | validation | error | qwen3-next, moe, shared-expert, fusion |
| ref2va requires at least one image reference | validation | error | minimax-h3, ref2va, missing-input |
| ref2va video preparation requires a video or video_audio… | validation | error | minimax-h3, ref2va, missing-input, not-implemented |
| reference audio sample rate must be positive | validation | error | minimax-h3, audio, sample-rate |
| Sampling factor should be a multiply of 2! | validation | error | subsampling, convolution, audio, config-validation, phi4 |
| sampling observer does not support pipeline-parallel… | exception | error | pipeline-parallel, sampling, observer, type-validation |
| --ssl-certfile requires --ssl-keyfile to be specified as… | validation | error | sglang, ssl, tls, server-config, argument-validation |
| The hpc_ops MoE runner backend does not support MoE GEMM… | validation | error | quantization, moe, moe-runner-backend, unsupported-feature |
| Tried to append None to state. | validation | error | none-check, operator-overload, dsl, validation |
| Unknown multimodal item type | validation | error | multimodal, routing, input-validation |
| Unsupported transport_mode | validation | error | configuration, transport-mode, env-var, receiver |
| `a`/`b` must be contiguous in the last dim. | validation | error | fla, fused-recurrent, contiguity, stride-check |
| A tokenizer is required to load an external ngram corpus. | validation | error | sglang, ngram, tokenizer, validation, null-argument |
| Bare IPv6 address without brackets is ambiguous | validation | error | network, ipv6, address-parsing, sglang |
| Block sparse tensors | validation | error | block-sparse, shape-mismatch, kv-length |
| cos and sin must have matching [S, D/2] shapes | validation | error | rope, cos-sin, shape-validation |
| head_dim must be positive, even, and <= 128 | validation | error | rope, head-dim, shape-validation, hunyuan |
| Invalid fused KV projection shape: got | validation | error | shape-validation, speculative-decoding, fused-kernel |
| kv-canary: read_bytes must be a multiple of | exception | error | kv-canary, alignment, validation, valueerror |
| kv-canary: RealKvSource.num_bytes_per_token must be a… | validation | error | kv-cache, alignment, byte-width, validation |
| --linear-attn-decode-backend flashkda is not supported… | validation | error | sglang, linear-attention, flashkda, prefill-only, server-args |
| MiniMax H3 request task must be a non-empty string | validation | error | minimax-h3, task-required, request-validation, sampling-params |
| nvfp4_gemm_swiglu_nvfp4_quant requires CUDA tensors | validation | error | nvfp4, cuda, device-placement, gpu-kernel |
| Only block_quant=True is supported in Quark MXFP4… | exception | error | quantization, fp8, block-quantization, quark, moe |
| Parameter not found in the model. | exception | error | quantization, bitsandbytes, parameter-mapping, checkpoint |
| SANA-WM streaming denoising expects 5D latents (B, C, T, H… | validation | error | sana-wm, streaming, offline, latent-shape, valueerror |
| `sigmas` and `timesteps` should have the same length | validation | error | scheduler, length-mismatch, validation |
| SSL certificate file not found | validation | error | sglang, ssl, tls, file-not-found, deployment |
| target.short_edge must be an integer, got | validation | error | minimax-h3, short-edge, type-error, spatial |
| task cannot carry image.target_canvas materials | validation | error | minimax-h3, task-validation, materials, plan-validation |
| Unexpected audio latents rank | exception | error | ltx-2, audio-latents, rank-check, video-generation |
| Unknown action domain name | validation | error | cosmos3, action-generation, domain-name, lookup, validation |
| VmmReservation.map_existing after close | exception | error | cuda, vmm, use-after-close, lifecycle |
| When additional_t_cond is True, addition_t_cond must be… | validation | error | qwen-image, timestep-conditioning, missing-argument |
| audio_out_channels must be divisible by tp_size for… | exception | error | tp-sharding, audio, divisibility, ltx-2, config-validation |
| browser.search requires a query | validation | error | browser, search, argument-validation, tool-calling |
| --enable-dsa-cache-layer-split is not supported on decode… | validation | error | dsa, pd-disaggregation, config-validation |
| expected mapping, got | validation | error | inkling, tool-calls, type-error |
| For Fused MoE layers, only | validation | error | quantization, moe, marlin, format, compressed-tensors |
| func_path should contain both module name and func name… | validation | error | dynamic-import, configuration, valueerror |
| GGUF models are not supported. | validation | error | model-format, gguf, unsupported |
| Hidden size must be divisible by num_heads | validation | error | config, validation, transformer, diffusion |
| Invalid precision: . Must be 'int4' or 'nvfp4 | validation | error | quantization, config-validation, nunchaku |
| min_new_tokens must be in | validation | error | sampling-params, validation, min-new-tokens, sglang |
| mm_process_config[' '] must be a dict, but got | validation | error | multimodal, json, server-args, type-error |
| Mobius fused gate/up destination is missing | exception | error | weight-loading, key-mapping, mobius, interns2 |
| model_index.json._minimax_h3.partition must be one of… | validation | critical | minimax-h3, model-index, partition, config-validation |
| pattern must contain at least one token | validation | error | sglang, tokens, pattern-matching, validation, kmp |
| PD Disaggregation does NOT support PD different TP sizes… | exception | error | pd-disagg, tp-degree-mismatch, hybrid-model, non-mla |
| Pipeline ' ' is already registered; pass overwrite=True to… | validation | error | registry, duplicate, value-error, pipeline, idempotency |
| positions must match ctx_hidden token count for fused KV… | validation | error | shape-validation, speculative-decoding |
| Transfer thread failed because of | exception | critical | pd-disagg, transfer-thread, mooncake, wrapper |
| unsupported MiniMax H3 audio material chain | validation | error | minimax-h3, material-chain, unsupported-operation |
| use_fast= conflicts with image_processor_backend= . | validation | error | multimodal, image-processor, conflicting-args |
| Block sparse tensors | validation | error | block-sparse, shape-mismatch, broadcasting |
| consumer_count must be positive | validation | error | memory-pool, config-validation, cuda-ipc, multimodal-transport |
| External corpus is empty — no tokens were loaded. | exception | error | ngram, corpus-loading, empty-input |
| generated MiniMax H3 MP4 has invalid size | exception | error | minimax-h3, mp4, dimensions, validation |
| head_dim must be a multiple of 8, got | validation | error | rope, model-config, shape-validation, ltx-2 |
| Intern-S2-Mobius does not support: " + "… | validation | error | model-support, pipeline-parallelism, expert-parallelism, config-validation |
| kitchen_w4a8 is inferred from per-layer checkpoint… | validation | error | quantization, api-misuse, w4a8 |
| kv-canary: write_offsets_len must equal write_req_capacity… | validation | error | kv-canary, offsets, off-by-one |
| manually start is only supported yet | exception | error | profiling, not-implemented, sglang |
| must be list or null; got | validation | error | validation, config, type-error |
| _block_cnt and _block_idx must both be provided or both be… | validation | error | block-sparse, attention, paired-arguments, validation |
| num_frames must be positive | validation | error | sana-video, video-generation, argument-validation, num-frames |
| Only one of `timesteps` or `sigmas` can be passed. Please… | validation | error | qwen-image, diffusers, scheduler, timesteps, sigmas, mutually-exclusive-args |
| Qwen3-Next MTP shared expert fusion currently supports… | validation | error | qwen3-next, mtp, shared-expert, speculative-decoding |
| reference image width and height must be positive finite… | validation | error | minimax-h3, image-shape, validation |
| SGLANG_INKLING_DEFAULT_REASONING_EFFORT must be numeric | exception | error | environment-variable, inkling, reasoning, server-config, sglang |
| sparse_mla_q8kv8_prefill_fwd supports d_qk=512/576, got | validation | error | shape-validation, mla, unsupported-dim |
| Unknown gemm type | validation | error | sglang, moe, humming, gemm, dispatch |
| Assistant tool call function.arguments must be a JSON… | validation | error | tool-calling, json, validation, openai-api, sglang |
| Block sparsity + paged KV not supported on SM100 | exception | error | flash-attention, block-sparse, paged-kv, sm100, unsupported-feature |
| Can't import trtllm_fp8_block_scale_routed_moe from… | exception | error | flashinfer, trtllm, moe, fp8, import-error, version-mismatch |
| cos/sin shape does not cover image tokens and head_dim | validation | error | rope, cos-sin, bounds-check, multimodal |
| Didn't find any LoRA adapters when trying to evict LRU LoRA… | validation | error | lora, eviction, capacity |
| Error: --model-type requires a value. | validation | error | rocm, allreduce, deterministic, alignment, bfloat16 |
| Expected a 2D, 3D, or 4D attention mask, got | validation | error | attention-mask, shape-validation, stable-diffusion |
| Failed to load the tokenizer. If you are using a LLaMA V1… | exception | error | tokenizer, huggingface, typeerror |