sgl-project/sglang
Documented errors, page 26 of 32. Back to sgl-project/sglang
| Code / Message | Type | Severity | Tags |
|---|---|---|---|
| keyframe denoising requires semantic_frame_indices in | validation | error | minimax-h3, keyframe, signature-validation |
| launch_local_runtime requires --dp-size 1; got dp_size= | validation | critical | startup, config, dp-size, parallelism |
| Load Intern-S2-Mobius through its conditional-generation… | exception | error | weight-loading, api-misuse |
| MHATokenToKOnlyPool does not allocate V | exception | error | kv-cache, k-only-pool, sparse-attention, minimax, sglang |
| --minimax-h3-adaln-online rebuilds AdaLN outputs from the… | validation | error | gguf, minimax-h3, adaln, config-conflict |
| MiniMax H3 ring size must be positive. | validation | critical | ring-attention, sequence-parallel, config, validation |
| miss_count must be int32. | validation | error | hisparse, dtype-validation, miss-plan |
| mlx_call_multi outputs must be MLX arrays | validation | error | sglang, mlx, type-check, callback, return-type |
| No free port available. | exception | error | network, port, test-infra |
| num_inference_steps must be positive, got | validation | error | sglang, ltx-2, num-inference-steps, sigma-schedule, diffusion, invalid-argument |
| NVIDIA KDA state must be [B,H,K,V] with (K,V)= | validation | error | sglang, nvidia, kda, tensor-shape, layout |
| Only ver=3 is supported for MUSA FA3. | validation | error | musa, flash-attention, moore-threads, sglang |
| q, k, and v must be CUDA tensors | validation | error | cuda, device, ulysses, attention |
| reference image target dimensions must be aligned to | validation | error | minimax-h3, alignment, resize |
| score_mod is only supported by the FA4 backend. | exception | error | score-mod, flash-attention-4, extend, sglang |
| does not support target_verify | exception | error | sglang, linear-attention, speculative-decoding, not-implemented, base-class |
| sglang.srt.layers.attention.nsa.tilelang_kernel is… | console | warning | deprecation, sglang, tilelang, import |
| sglang.srt.layers.attention.nsa.utils is deprecated; use… | console | warning | deprecation, sglang, utils, import |
| top_p must be scalar or have one value per row, got | validation | error | sampling, top-p, batch-size-mismatch, validation |
| Unexpected RoPE rank | validation | error | sana-wm, rope, tensor-rank, refiner |
| W4AFP8 group_size must be a positive integer, got | validation | error | humming, group-size, config-validation, quantization |
| waveform must be a 1D tensor, but got | validation | error | multimodal, audio, tensor-shape, mimo |
| weight cache transport backend | exception | error | transport, not-implemented, weight-cache, config |
| Adapter not found in the pipeline. Please provide lora_path… | validation | error | lora, missing-adapter |
| Audio placeholder count does not match audio_data | validation | error | multimodal, placeholder-mismatch, audio, valueerror |
| Batch tokenization is not needed for input_embeds. Do not… | validation | error | sglang, config, input-embeds, batch-encode |
| browser.find requires a pattern | validation | error | browser, find, argument-validation, tool-calling |
| Conflicting kernel registration for op | validation | error | kernel-registry, duplicate-registration |
| Cosmos3 prompt batch must not be empty | validation | error | cosmos3, tokenization, empty-prompt, empty-batch |
| domain_id must be non-negative, got | validation | error | cosmos3, action-generation, domain-id, validation |
| Duplicate request ID detected | validation | error | request-id, duplicate, retry, batching |
| Grammar mask max_rows must be positive, got | exception | error | grammar, constrained-decoding, validation |
| kv must be contiguous | validation | error | contiguity, kv-cache, cuda |
| LTX-2 requires audio latents for denoising. | exception | critical | ltx-2, audio-latents, required-input |
| MiniMax H3 image material is invalid | validation | error | minimax-h3, image-decode, pil, corrupt-file |
| No image data in response | validation | error | comfyui, empty-response, base64, image-generation |
| pair_postprocess must return the same shape as input | validation | error | scheduler, tensor-shape, callback |
| --prefill-only-disable-kv-cache requires… | validation | error | sglang, radix-cache, prefix-cache, kv-cache, server-args |
| Prompt cannot be empty | validation | error | validation, comfyui, empty-prompt, image-generation |
| SANA-WM refiner expects 5D latents shaped (B, C, T, H, W)… | validation | error | sana-wm, refiner, latent-shape, ndim, valueerror |
| --sidecar-args must be a JSON array of strings. | validation | error | grpc, sidecar, json-validation, config-validation |
| SM120 relative bias currently supports head_dim and… | validation | error | flash-attention, sm120, relative-bias, head-dim |
| sparse_attn_v4_paged_prefill expects fp16/bf16 q, got | validation | error | attention, dtype, triton, prefill |
| SSL key file not found | validation | error | sglang, ssl, tls, file-not-found, deployment |
| st attn not supported | validation | error | sliding-tile-attention, missing-dependency, native-extension, init |
| task alias targets undeclared task | validation | error | minimax-h3, task-aliases, dangling-reference, model-index |
| token_ids_logprob contains out-of-vocabulary token id | validation | error | sglang, vocab, out-of-range, logprob |
| torchaudio is required for audio inputs; install torchaudio | error_code | error | multimodal, audio, missing-dependency, mimo |
| Unsupported cute dtype | validation | error | flashinfer, cutedsl, moe, unsupported-dtype |
| Unsupported layout for models with head_dim != v_head_dim… | validation | error | sglang, mla, hicache, io-backend, layout |
| Unsupported module payload type for load_format= | validation | error | weights-update, payload-type, validation, multimodal |
| unsupported task alias mapping | validation | error | minimax-h3, task-aliases, canonicalization, model-index |
| Decode context parallel size (--dcp-size /… | validation | error | server-args, dcp, parallelism, validation, sglang |
| DSpark with dp attention does not support context parallel | validation | error | speculative-decoding, dspark, context-parallel, dp-attention |
| kv-canary: RealKvSource.tensor must be at least 2-D, got… | validation | error | kv-cache, shape-validation, tensor-rank |
| kv-canary: scatter_req_token_ids pool_out must be 2-D, got… | validation | error | kv-cache, shape-validation, tensor-rank |
| kv-canary: . already wrapped by kv-canary | exception | error | kv-canary, monkey-patching, idempotency, runtimeerror |
| --minimax-h3-adaln-cache-path requires the unquantized… | validation | error | gguf, minimax-h3, adaln-cache, config-conflict |
| Missing required keys in config | validation | error | hf3fs, config-validation, missing-keys |
| MLX 0.32 does not support complex128; convert the Torch… | validation | error | sglang, mlx, dtype, complex-numbers, torch |
| MLX float64 arrays cannot be exported to a Torch MPS… | validation | error | sglang, mlx, float64, mps, dtype |
| MlxTensorView requires a Torch MPS tensor, got | validation | error | sglang, mlx, mps, device, torch |
| model_index.json._minimax_h3 must be an object | validation | critical | minimax-h3, model-index, config-validation |
| port is unavailable and --strict-ports is enabled. Either… | exception | critical | network, ports, startup, config |
| NVFP4 embedding is gather-only. Reaching here means a tied… | exception | error | |
| Serialized W4A8 checkpoints are not supported on MPS | validation | critical | quantization, mps, platform-support, w4a8 |
| think_excluded_token | validation | error | reasoning, tokenizer, grammar |
| tiny_gemm: no valid split_n for N= | error_code | error | gemm, tiny-gemm, split-n, block-size |
| Unknown flash attention version | validation | error | flash-attention, version-dispatch, attention, invalid-argument |
| unsupported content part type | validation | error | inkling, content-parts, unsupported-type |
| use_rope2d must be True | validation | error | step3-vl, vision-config, rope |
| composite_input event payload requires non-empty input_types | validation | error | realtime, event-validation, composite-input, missing-field |
| Connection closed while reading message body | error_code | error | protocol, socket, eof, truncated |
| Cosmos3 observation image arrays must use uint8 dtype | validation | error | cosmos3, numpy, dtype, image-input |
| Failed to download Real-ESRGAN weights from HuggingFace repo | exception | error | realesrgan, huggingface, download, postprocess |
| Failed to register buffer to Mooncake Store, error code | exception | critical | mooncake, rdma, memory-registration, transfer-engine |
| hd256 forward varlen expects q rank 3 or 5, got rank | exception | error | cuda, attention, tensor-rank, varlen, head-dim-256 |
| Invalid external ngram corpus record at line | validation | error | sglang, ngram, json, jsonl, schema-validation |
| mask_strategy cannot be None for SlidingTileAttention | validation | error | sliding-tile-attention, mask-strategy, none-check, forward |
| q must be torch.float8_e4m3fn, got | validation | error | dtype, fp8, sparse-attention, sglang |
| row | validation | error | validation, config, nested-list |
| Unsupported interleaved_rope_fp64 dtype | validation | error | jit, dtype, bfloat16, rope |
| Unsupported Spark2_5 layer_type | validation | error | spark2-5, layer-type, config |
| Video generation timed out after | error_code | error | timeout, video-generation, polling, sgldiffusion |
| cos and sin must be CUDA and last-dim contiguous | validation | error | rope, cuda, device-placement, contiguity |
| expected one fd, got | exception | error | ipc, fd-passing, protocol |
| Inkling thinking part payload must be a string | validation | error | inkling, thinking, type-error |
| Invalid threshold_type for topk | validation | error | validation, value-error, moba, attention, config |
| labels() cannot be called on an already-labeled metric. | validation | error | observability, ray, metrics, api-misuse |
| MiniMax H3 num_steps must be > 0 | validation | error | minimax-h3, sampling, num-steps, invalid-argument |
| must be a tensor, list of tensors, or None. | validation | error | batching, type-error, validation |
| must be on device , got | validation | error | device-validation, multi-gpu, output-buffer, sparse-mla |
| must be on the same device as block sparse tensors | validation | error | block-sparse, device-mismatch, metadata |
| Nunchaku SVDQuant is only supported on NVIDIA CUDA GPUs… | validation | error | nunchaku, svdquant, cuda, platform-support, quantization |
| must be a dict or a comma-separated component=value string | validation | error | config-parsing, component-map, type-validation |
| seed must not exceed the signed int64 maximum, got | validation | error | minimax-h3, seed, int64-overflow, validation |
| is not supported. Please make sure to choose one of… | validation | error | scheduler, diffusion, timesteps, unipc |
| timed out waiting for POSIX fd exchange | exception | error | distributed, timeout, unix-socket, fd-passing |
| Unexpected a shape for varlen | validation | error | kda, varlen, shape-validation, cutedsl |
| Unsupported topk_mode | validation | error | flashvdm, topk, hunyuan3d, config, value-error |