sgl-project/sglang
Documented errors, page 15 of 32. Back to sgl-project/sglang
| Code / Message | Type | Severity | Tags |
|---|---|---|---|
| FlexKV layerwise worker NACK'd eventfd transfer | exception | critical | flexkv, eventfd, nack, file-descriptors |
| hicache_host_memory_mode must be 'cache' or 'buffer_only'… | validation | error | sglang, hicache, host-memory, config-validation, server-args |
| kv-canary: expected input tensors must be None when… | validation | error | kv-canary, argument-validation, mutually-exclusive |
| LoRA is only compatible with NGRAM, EAGLE, NEXTN, EAGLE3… | validation | error | sglang, lora, speculative-decoding, feature-incompatibility |
| must be True / False / str / list[str], got | validation | error | decorator, distributed, type-validation, rank-consensus, sglang |
| No striding allowed for non-symmetric convolutions! | validation | error | convolution, causal, stride, padding, phi4 |
| Not support activation_func | exception | error | kimi-k3, vision, activation, config-validation |
| Remote media exceeds the | validation | error | media, size-limit, download, security |
| SGLang's AutoRound GPTQ loader supports desc_act=False… | validation | error | quantization, auto-round, gptq, desc-act |
| Some weights are not initialized from checkpoints | error_code | critical | checkpoint-loading, weights, model-init, deepseek-ocr |
| Staging is enabled but kv_manager.kv_buffer_tensors is… | exception | error | disaggregation, staging-buffer, kv-buffer-tensors, initialization |
| The hpc_ops MoE runner backend does not support… | validation | error | sglang, moe, hpc-ops, router-weight, config-validation |
| Unknown quantization strategy | validation | error | quantization, fp8, w8a8, strategy, post-load |
| assistant message cannot mix reasoning_content with ordered… | validation | error | inkling, reasoning, conflicting-fields |
| does not support an explicit quantization override; use a… | validation | error | quantization, server-args, config, model-loading |
| Cosmos3 inverse_dynamics input requires an observation video | validation | error | cosmos3, inverse-dynamics, missing-video, input-validation |
| cu_seqlens_q and cu_seqlens_k must describe the same batch | validation | error | npu, ascend, varlen, batch-mismatch, validation |
| CUDA VMM multimodal pool is closing | exception | error | cuda, vmm, shutdown, race-condition, lifecycle |
| Failed to generate video | exception | error | comfyui, video-generation, error-wrapping, network |
| ffprobe failed for final MiniMax H3 output | exception | error | minimax-h3, ffprobe, corrupt-output, validation |
| --fp8-gemm-backend=deep_gemm cannot serve MXFP8 weight shape | exception | error | quantization, mxfp8, deep-gemm, hardware-compatibility, gemm-backend |
| HiSparse speculative swap requires 2-4 steps, got | validation | error | hisparse, speculative-decoding, shape-validation |
| Invalid grid metadata for kimi image tokens | validation | error | kimi, multimodal, grid-metadata, validation |
| Invalid JSON in external ngram corpus at line | validation | error | sglang, ngram, json, jsonl, corrupt-data |
| Invalid value | validation | error | operator-overload, fork-join, type-error, dsl |
| KDA `dt_bias` must be a contiguous 1D or 2D tensor. | exception | error | kda, replayssm, contiguity, parameter-shape |
| : expected , got | validation | error | config, type-coercion, validation |
| Kimi expert-pack identity mismatch at index | validation | critical | kimi, moe, expert-pack, index-integrity |
| LTX2 stage-1 CFG parallel degree exceeds guidance pass… | exception | error | ltx-2, cfg-parallel, config-mismatch, distributed |
| `mu` must be passed when `use_dynamic_shifting` is set to… | validation | error | scheduler, diffusion, missing-parameter, dynamic-shifting |
| prediction_type given as | validation | error | scheduler, diffusion, prediction-type, unipc |
| tar material offset_data and size must be non-negative | validation | error | minimax-h3, tar, material-uri, validation |
| The hpc_ops MoE runner backend only supports FP8-quantized… | validation | error | sglang, moe, hpc-ops, fp8, quantization, config-validation |
| consumer rank is outside | error_code | critical | distributed, rank-validation, cuda-ipc, multimodal-transport |
| Unknown cache_dit_params keys | validation | error | cache-dit, unknown-key, request-validation, config |
| Unsupported Quanto weight type for | validation | error | quantization, quanto, int8, unsupported-dtype |
| Z-Image expects one caption embedding per image, got | validation | error | z-image, batching, shape-validation |
| batch_draft_token_num config value | exception | error | config-validation, ngram, speculative-decoding, constructor |
| condition_video_keep must be 'first' or 'last', got | validation | error | cosmos3, v2v, condition-video-keep, enum-validation |
| Dots SWA latent decode requires page_size=64, got | exception | error | dots-hybrid, swa-mla, page-size, attention-backend, sglang |
| Duplicate request_id | validation | error | disaggregation, request-state, duplicate-id, idempotency |
| Invalid expert_pack configuration:\n | validation | error | gguf, expert-pack, sglang, path-validation |
| Invalid prompts type for score_prompts. | validation | error | scoring, type-validation, input-format |
| mean is more than 2 std from [a, b] in… | console | warning | pytorch, initialization, numerics, trunc-normal |
| num_heads must be divisible by num_epi_subtiles | validation | error | cutedsl, kernel-config, shape-validation, attention |
| --prefill-only-disable-kv-cache is not supported for | exception | error | prefill-only, mamba, fp4-kv, kv-cache |
| {reason} | validation | error | sglang, kv-events, endpoint-mismatch, dp-size |
| Resolved LoRA weight | validation | error | lora, download, cache |
| SANA-WM refiner decoding expected a sink frame plus refined… | validation | error | sana-wm, refiner, decode, temporal-length, valueerror |
| sendmsg sent bytes, expected | exception | error | ipc, fd-passing, sendmsg, weight-cache |
| SGLANG_RUST_SERVER serves the PD KV bootstrap registry on… | validation | error | sglang, pd-disaggregation, rust-server, bootstrap-port, port-conflict |
| STANDALONE speculative decoding requires the draft model to… | validation | critical | speculative-decoding, vocabulary-mismatch, model-config |
| Unknown pair_postprocess name | validation | error | scheduler, invalid-name, enumeration |
| AITER Sage backend does not have a metadata builder. | exception | error | aiter, sage-attention, attention-backend, rocm, not-implemented |
| attn_sink requires topk_length to be provided as well | validation | error | argument-validation, attention-sink, sparse-attention |
| Can't get gguf config for | exception | error | gguf, config, unsupported-architecture |
| DSV4 ragged verify does not support online c128 MTP; set… | exception | critical | deepseek-v4, ragged-verify, mtp, online-compress, env-var, sglang |
| External corpus ' ' already exists. Remove it before adding… | exception | error | ngram, corpus-loading, duplicate-key |
| head_dim mismatch across layers for fused KV path: expected | validation | error | config-validation, attention |
| initialize opentelemetry error | error_code | error | observability, tracing, opentelemetry, otlp-endpoint |
| Invalid direction | validation | error | enum-value, routing, dual-tower, forward-pass |
| Invalid quantization method | validation | error | quantization, configuration, startup, sglang |
| Invalid X-Data-Parallel-Rank header: must be an integer, got | http | error | http-header, data-parallel, routing, sglang |
| LingBotVideoBlock expects token-level temb6 with shape… | exception | error | shape-validation, timestep-embedding, lingbot, video-diffusion |
| MiniMax H3 AdaLN cache model_variant does not match the… | validation | error | minimax-h3, adaln-cache, variant-mismatch |
| must have dtype , got | validation | error | dtype-validation, output-buffer, sparse-mla, fp8 |
| num_heads ( ) must be divisible by ulysses_degree ( ). | exception | critical | parallelism, ulysses, attention-heads, lingbot, config-validation |
| O tensor must match dtype | validation | error | flash-attention, dtype-mismatch, cuda-kernel, validation |
| out, max_logits and lse must not alias each other | validation | error | mla, sparse-attention, buffer-aliasing, output-buffers |
| prefetch_timeout_base must be number, got | validation | error | hicache, config-validation, prefetch, type-error |
| prefetch_timeout_max must be number, got | validation | error | hicache, config-validation, prefetch, type-error |
| RayPrometheusMetric requires Ray to be installed. Install… | exception | error | observability, ray, metrics, missing-dependency |
| cannot transport an empty tensor | validation | error | multimodal, empty-tensor, validation |
| Unable to find matching target for | validation | error | quantization, compressed-tensors, layer-matching, config-mismatch |
| unsupported capture mode | validation | error | sana-wm, streaming-refiner, kv-capture, mode-string, valueerror |
| Unsupported layout for models with head_dim != v_head_dim | validation | error | sglang, mla, kv-cache, layout, host-pool |
| Unsupported video input type | validation | error | video, multimodal, input-validation, valueerror |
| Cosmos3CausalAttention requires num_attention_heads… | validation | error | sglang, cosmos3, tensor-parallel, attention-heads, config |
| does not support QVG KV-cache quantization | validation | error | kv-cache, quantization, unsupported-feature |
| duplicate fd for | exception | error | distributed, fd-passing, duplicate-key, protocol |
| Failed to create shm file | exception | critical | shm, tmpfs, disk-full, hicache, docker |
| requires to be resident; got from --component-residency | validation | error | config, residency, feature-conflict |
| {finish_reason["message"]} | exception | error | sglang, abort, bad-request, non-streaming |
| FlashAttention-4 CUTE is not available. Install… | exception | error | flash-attention, fa4, sm120, import-error, missing-dependency |
| head_first is deprecated and will be removed in a future… | exception | warning | fla, gated-delta-rule, deprecation, tensor-layout |
| model_index.json does not contain _diffusers_version | validation | error | diffusers, config-validation |
| Module instance is not unique | validation | error | nvtx, pytorch-hooks, shared-weights, profiling, sglang |
| No embedding available for request | http | error | internal, encoder, state-machine, race-condition |
| NPU packed attention requires q, k, and v on the same NPU… | validation | error | npu, ascend, device-placement, validation |
| nvfp4_gemm_swiglu_nvfp4_quant currently supports NVFP4… | validation | error | nvfp4, dtype-validation, quantization |
| opentelemetry package is not installed!!! Please not enable… | error_code | error | observability, tracing, opentelemetry, missing-dependency |
| rids_to_check cannot be used in PP mode | validation | error | disaggregation, pipeline-parallel, api-misuse, argument-conflict |
| Serve backend uses API version ; this SGLang release… | exception | critical | sglang, cli, version-mismatch, plugin-abi |
| --ssl-keyfile-password has no effect without --ssl-certfile… | validation | error | sglang, ssl, tls, secrets, argument-validation |
| The requested SM120 sheared-bias specialization exceeds… | validation | error | flash-attention, sm120, shared-memory, relative-bias |
| top_p values must be in (0, 1] | validation | error | sampling, top-p, value-range, validation |
| Unknown cache_dit_params['secondary'] keys | validation | error | cache-dit, secondary-cache, unknown-key, request-validation |
| Unknown output type | exception | error | harmony, output-parsing, type-dispatch, sglang |
| Video generation failed | error_code | error | video-generation, server-side-failure, polling, sgldiffusion |
| A CUDA VMM-enabled model must provide a multimodal processor | exception | error | cuda, vmm, multimodal, config |