sgl-project/sglang
Documented errors, page 5 of 32. Back to sgl-project/sglang
| Code / Message | Type | Severity | Tags |
|---|---|---|---|
| prefetch_threshold must be int, got | validation | error | hicache, config-validation, prefetch, type-error |
| ResolvedPlan requires one or two ordered image/keyframe… | validation | error | minimax-h3, keyframe, frame-index, plan-validation |
| time must have shape [batch] | validation | error | pi05, tensor-rank, input-validation |
| Unsupported dots note omni video_config fields | validation | error | multimodal, video-config, unknown-fields, valueerror |
| Unsupported preprocessed video item | validation | error | multimodal, video, schema-mismatch, version-drift |
| DeepEP v2 MoE is not validated as a speculative draft… | validation | error | speculative-decoding, deepep, moe, draft-model, server-args |
| DSPARK aux hidden capture requires PP=1. | exception | error | dspark, pipeline-parallel, speculative-decoding, kimi |
| Failed to render chat template for embedding input | validation | error | jinja, embeddings, chat-template, type-error |
| `g_cache` must have dtype torch.float32. | exception | error | helion, kda, replayssm, dtype, gate-cache |
| generated MiniMax H3 MP4 size does not match the resolved… | exception | error | minimax-h3, resolution, geometry-mismatch, validation |
| Inkling shared-sink LoRA outer factors must have expert… | validation | error | sglang, lora, shape-validation, moe, shared-experts, inkling |
| Model architectures failed to be inspected. Please check… | validation | error | registry, model-resolution, architecture-unsupported |
| RaggedVerifyLayout requires at least one request | validation | error | speculative-decoding, validation, batch-empty |
| refiner cu_seqlens live text length must be in | validation | error | minimax-h3, dit, shape-validation, text-embeddings |
| Unknown MLX async mode | validation | error | mlx, async-scheduler, internal-contract, sglang |
| Weight input_size_per_partition = | validation | error | marlin, gptq, tensor-parallel, shape-validation |
| Cannot resolve the consumer rank before parallel state… | exception | critical | distributed, parallel-state, cuda-ipc, initialization-order |
| Comfy INT8 layer needs I8 weights and F32 scales, got and | validation | error | quantization, int8, dtype-mismatch, checkpoint-validation |
| Cosmos3 forward_dynamics produces video; use /v1/videos… | validation | error | cosmos3, action-mode, endpoint-routing, video-generation |
| FlashAttention combine kernel cannot be implemented with… | error_code | critical | cuda, attention, split-kv, kernel-config |
| No frames before start_time | validation | error | video, timestamps, segment, out-of-range |
| RPC error on | exception | error | rpc, debug-utils, remote-error, sglang |
| The argument disaggregation-decode-enable-offload-kvcache… | validation | error | sglang, pd-disaggregation, kv-offload, decode-only, config-validation |
| Unsupported activation | validation | error | exaone, activation, model-config, unsupported-operation |
| Cosmos3 batched prompts must tokenize to the same length… | validation | error | cosmos3, multimodal-gen, batching, tokenization, validation |
| cuMemGetAddressRange | exception | error | cuda, vmm, cuda-graph, invalid-pointer |
| --dcp-replicate-q-proj requires --dcp-size > 1. | validation | error | sglang, distributed, dcp, argument-validation |
| --enable-tp-lm-head-all-to-all requires an available PyNCCL… | validation | critical | distributed, nccl, p2p, tp, lm-head |
| Hugging Face weight URL has no filename | validation | error | huggingface, weights, url-parsing |
| Multimodal data is corrupted or cannot be decoded | exception | error | multimodal, wrapper, decode, request |
| QwenImage text conditioning mask has shape | validation | error | qwen-image, shape-mismatch, mask-validation, text-embedding |
| [Staging] KV transfer via staging buffer failed | exception | error | disaggregation, staging-buffer, transfer-failure, wrapper-exception |
| Action endpoint requires SamplingParams or… | validation | error | action-endpoint, sampling-params, model-registry, subclass-validation |
| conflicting safetensors LoRA alpha metadata | validation | error | lora, safetensors, metadata, conflict |
| --enable-int8-mamba-checkpoint is not supported together… | validation | error | sglang, mamba, int8, hierarchical-cache, incompatible-flags, server-args |
| {failure_msg}: {error_msg} | exception | error | lora, scheduler, adapter-loading, runtime-error |
| Humming quantization for MoE only supports… | validation | error | moe, humming, runner-backend, config-validation |
| Ideogram4DenoisingStage applies its custom scheduler step | exception | error | ideogram, scheduler, unsupported-operation, diffusers |
| invalid compressor checkpoint name | validation | error | gguf, deepseek, compressor, name-mapping |
| Invalid config pair (missing '=') | validation | error | config, validation, debug-utils |
| Invalid latent H/W computed from batch.height/width | validation | error | ltx-2, video-generation, resolution-validation, sequence-parallelism |
| NPU packed attention does not support a sequence that is… | exception | error | npu, ascend, varlen, ring-attention, not-implemented |
| SANA-WM refiner text encoder must return per-layer… | exception | error | sana-wm, text-encoder, hidden-states, runtimeerror, model-output |
| top_logprobs_num exceeds disaggregation metadata capacity … | validation | error | sglang, disaggregation, logprobs, metadata-buffer, capacity |
| Unsupported quantized embedding marker for | validation | critical | quantization, checkpoint, embedding, model-load |
| You have passed a list of generators of length | validation | error | mova, generator, batch-size, diffusers, validation |
| All memory audio latents must share batch and channel… | validation | error | joy-echo, audio, latent, memory-bank, concat |
| All ranges must be within | validation | error | attention, varlen, mask-metadata, range-validation, multimodal |
| config.json claims a packaged Muse Glimmer MLX artifact | validation | error | checkpoint-format, packaging, weight-keys |
| flash-attn is not installed. Please install it, e.g., `pip… | exception | critical | flash-attn, import-error, dependency-missing, attention, cuda |
| Speculative decoding with --enable-unified-memory is only… | validation | error | speculative-decoding, unified-memory, hybrid-swa, kv-cache |
| Unknown data: . You may need to set `--data-type` if using… | exception | error | data-schema, polars, cli, text-comparison |
| Comfy NVFP4 layer needs a 2D packed weight, got | validation | error | quantization, nvfp4, rank-mismatch, conv-weights, checkpoint-validation |
| config not published; cannot read a config leaf | validation | error | config, runtime-context, initialization-order, sglang |
| Eagle3 MLA draft post_load_weights only supports float… | exception | error | eagle3, weight-loading, dtype, quantization, speculative-decoding |
| --enable-metrics requires smg-grpc-servicer ≥ 0.5.3 (the… | exception | error | grpc, metrics, version-mismatch, dependency, sglang |
| `fuse_qkv_projections()` is not supported for models having… | validation | error | fusion, attention, optimization, value-error |
| Grouped pipeline returned fewer outputs than requests. | error_code | critical | runtime, pipeline, inference, internal-error, multimodal |
| MiniMax-H3 adaln_t_table must have shape [N, D] with N >=… | validation | error | minimax-h3, safetensors, checkpoint, tensor-shape, validation |
| MiniMaxH3TextEncodingStage direct Qwen3VL encoder forward… | validation | error | minimax-h3, legacy-api, not-implemented, request-format |
| PD disaggregation for MiniMax sparse layers with index… | validation | error | disaggregation, minimax, sparse-attention, not-implemented |
| SGLANG_ENABLE_EPLB_BALANCEDNESS_METRIC is no longer… | validation | error | env-var, eplb, metrics, migration, server-args, deprecated |
| task does not allow condition role= type= | validation | error | minimax-h3, condition-rules, task-profile, validation |
| tokenizer missing required special token | exception | critical | tokenizer, checkpoint-mismatch, vocab, asr, initialization |
| unsupported input for Sana fused bias-GLU | exception | error | sana, diffusion, glu, channels-last, triton |
| Cannot call `set_default_attn_processor` when attention… | validation | error | attention, processor, flux2, state-error |
| Cannot resolve total_kv_heads: kv_args has neither… | exception | error | disaggregation, staging-buffer, kv-heads, missing-metadata |
| content hash mismatch for media_data | validation | error | multimodal, artifact-cache, content-hash, cache-invalidation |
| DFLASH sliding_attention layers require… | validation | error | speculative-decoding, dflash, sliding-window, hf-config |
| --diff-threshold with a single argument must be a float… | validation | error | cli, parsing, float, threshold |
| H3 conditioning projection | validation | critical | minimax-h3, conditioning-projection, shape-mismatch, checkpoint |
| Missing config.sliding_window for Mellum sliding_attention… | exception | critical | mellum, sliding-window, config-validation |
| Missing rope_parameters | exception | critical | mellum, config-validation, rope, model-loading |
| q shape must be [num_tokens, num_qo_heads, head_dim], got | validation | error | shape-validation, rope, gqa, metal, sgl-kernel |
| quantized tensor maps to a non-weight parameter | validation | error | gguf, deepseek, quantization, weight-mapping |
| SGLang does not recognize target_modules= | validation | error | lora, target-modules, peft-config, sglang |
| mask/position shape mismatch: vs . | exception | error | sglang, speculative-decoding, dflash, internal-invariant, mask-shape |
| DFLASH config.layer_types must be a sequence of attention… | validation | error | speculative-decoding, dflash, hf-config, type-validation |
| Language ' ' not recognized. Use full name (e.g.… | exception | error | whisper, audio, language-code, validation |
| must use component=value entries | validation | error | config, cli, validation, layerwise-offload |
| rmsnorm_hf: unsupported hidden_size= | validation | error | rmsnorm, shape-validation, cuda-kernel, layernorm |
| The batch size is expected to be 1 rather than | validation | error | kda, fla, varlen, batch-shape, attention |
| action policy returned no output | exception | error | action-inference, empty-output, scheduler |
| Cosmos3 AVAE dec_strides product must equal hop_size… | validation | error | avae, config, cosmos3, init, value-error |
| --enable-deepseek-v4-fp4-indexer requires SM100 or SM120… | validation | error | sglang, deepseek, fp4, gpu-architecture, deepgemm |
| Hf3fsClient.check | validation | error | hf3fs, validation, alignment, batch-io |
| Language ' ' is not in this Whisper model's vocabulary. The… | exception | error | whisper, audio, tokenizer-vocab, model-version |
| MiniMax H3 loaded checkpoint partition does not match… | validation | error | minimax-h3, checkpoint-mismatch, model-variant, integrity-check, sglang |
| MIXED_PRECISION layer group | exception | error | quantization, mixed-precision, quark, unsupported-algo, mxfp4, python |
| MXFP8 dense GEMM requested via… | error_code | error | quantization, mxfp8, gemm-backend, flashinfer, hardware-compatibility |
| no supported CUDA VMM allocation handle type | exception | critical | cuda, vmm, handle, driver-capability, environment |
| Reasoning parser ' ' is always-on and cannot be disabled… | validation | error | anthropic-api, reasoning, toggle |
| Unsupported Pi05 dtype | validation | error | pi05, dtype, checkpoint, config |
| byte_tensor must be a sufficiently large contiguous uint8… | validation | error | multimodal, cuda, tensor-validation, memory-pool |
| config not published for role | exception | error | runtime-context, initialization-order, roles, sglang |
| NIXL memory registration failed for aux tensors | exception | critical | nixl, memory-registration, startup, disaggregation |
| op has no backend usable on device (registered: ) | validation | error | kernels, device-eligibility, environment, missing-dependency, sglang |
| PD state transfer failed: unknown state_type= | exception | error | disaggregation, state-transfer, dispatch, unsupported-feature |
| The block-wise quantization only supports dynamic… | validation | error | quantization, fp8, block-quant, activation-scheme |
| The block-wise quantization only supports fp8-serialized… | validation | error | quantization, fp8, block-quant, checkpoint-config |