sgl-project/sglang

Documented errors, page 23 of 32. Back to sgl-project/sglang

Code / Message	Type	Severity	Tags
Kimi expert-pack index SHA-256 does not match manifest	validation	critical	kimi, moe, expert-pack, sha256, integrity
manual_divisions must have	validation	error	pdmux, config, partitioning
Mooncake transfer_sync failed for	http	critical	mooncake, rdma, transfer, network
must be a CUDA tensor	validation	error	mla, quantization, cuda-device, scale-validation
No frames decoded from video	validation	error	cosmos3, v2v, video-decode, ffmpeg, corrupt-file
[Staging] Bulk RDMA transfer failed with ret=	exception	critical	mooncake, rdma, kv-cache, staging
The option is not yet implemented	console	warning	cli, not-implemented, argparse
thinking.display is not allowed when thinking.type is…	validation	error	anthropic, thinking, display, validation, request-validation
Unexpected dt_bias shape	validation	error	kda, linear-attention, shape-validation, cutedsl
Unknown interaction strategy	validation	error	config-validation, enum-value, dual-tower, mova
hidden_size must be positive.	validation	critical	config, model-architecture, validation, minimax-h3
Invalid k_out device/dtype for fused KV materialization…	validation	error	device-dtype-validation, kv-cache
Kimi expert-pack index is truncated	validation	critical	kimi, moe, expert-pack, truncated-index
Kimi expert-pack SHA-256 does not match manifest	validation	critical	kimi, moe, expert-pack, sha256, integrity
kv-canary: scatter_req_token_ids flat_in must be 1-D, got…	validation	error	kv-cache, shape-validation, tensor-rank
declares python-module but must also set `package.name` and…	validation	error	rust, cargo, manifest, validation
MiniMax H3 shift_scale must be > 0	validation	error	minimax-h3, sampling, shift-scale, invalid-argument
num_kv_heads mismatch across layers for fused KV path…	validation	error	config-validation, gqa, speculative-decoding
q, k, and v must be contiguous in head_size	validation	error	stride, contiguity, ulysses, triton
Quantization method specified in the model config	exception	error	quantization, config-mismatch, cli-args, sglang
return_hidden_states must be a boolean or the string…	validation	error	sglang, hidden-states, enum-validation, input-validation
SANA-WM forward requires encoder_hidden_states.	validation	error	sana-wm, missing-argument, required-parameter
Server process exited	exception	critical	server, lifecycle, oom, subprocess
Step3-VL image item has num_patches > 0 but no…	validation	error	step3-vl-10b, multimodal, patches
The input_ids contains values greater than the vocab size (…	validation	error	sglang, input-ids, vocab, batch
The layout of mBias is wrong	error_code	error	flash-attention, tensor-layout, bias, sm100
trtllm_mha backend can only be used with non-MLA models.	validation	error	attention-backend, trtllm-mha, mla, model-arch-mismatch, sglang
Unknown role	validation	error	chat-api, role-validation
Unrecognized image input, support local path, http url…	validation	error	image, loading, base64, input-validation
Attention backend name must be a string	validation	error	attention-backend, type-validation, config
CUDA VMM multimodal transport selected POSIX_FD, but this…	exception	critical	cuda, vmm, fabric, gpu-topology, runtimeerror
Current platform does not support NVFP4 quantization…	validation	error	nvfp4, gpu-hardware, blackwell, quantization, moe
--enable-dsa-cache-layer-split currently only supports the…	validation	error	dsa, pd-disaggregation, transfer-backend, config-validation
Kimi expert-pack physical role order is unsupported	validation	critical	kimi, moe, expert-pack, role-order
kv-canary: pool_slot_count must be positive, got	exception	error	kv-canary, validation, kv-cache-pool, value-error
MiniMax-H3 checkpoint shards disagree on adaln_t_table shape	validation	error	minimax-h3, safetensors, multi-shard, shape-mismatch
`mixed_qkv` must be 2D	validation	error	pytorch, tensor-shape, replayssm, decode, validation
MooncakeStore with standalone_storage=True requires…	exception	error	mooncake, allocator, standalone-storage, config
SANA-WM realtime denoising requires a realtime session	validation	error	sana-wm, streaming, session, realtime, valueerror
pool slot generation exhausted	error_code	error	multimodal, memory-pool, counter-overflow, long-running
Serialized W4A4 layer	validation	critical	quantization, shape-mismatch, w4a4
--sidecar requires --grpc-port or SGLANG_GRPC_PORT.	validation	error	grpc, sidecar, config-validation
SM120 relative bias requires tile_mn=(64, 128)	validation	error	flash-attention, sm120, tile-config, relative-bias
The following requested LoRA adapters are not loaded	validation	error	lora, registry, not-found, request, sglang
Unknown browser action	validation	error	browser, unknown-action, tool-calling, dispatch
Validate failed: S( ) must be divisible by F( ).	validation	error	shape, video-diffusion, modulation, validation
Apple Metal shader compiler not found. Install a full Xcode…	console	error	metal, build, xcode, shader-compiler, macos
camera_actions event payload must be list[list[str]]	validation	error	realtime, event-validation, camera-actions, type-mismatch
D= not supported, must be multiple of 256 and <= 8192	validation	error	shape-constraint, cuda-kernel, diffusion, norm
Either master_server_address or client_server_address is…	validation	error	config, mooncake, validation, missing-field
expert-pack v1 supports only single-GPU TP=EP=1	validation	error	expert-pack, moe, tensor-parallel, single-gpu
Hunyuan3D only supports num_outputs_per_prompt=1.	validation	error	hunyuan3d, input-validation, unsupported-parameter
indices must have shape (s_q, h_kv, topk), got	validation	error	rank-validation, indices, sparse-mla, topk
transition must be a map	validation	error	realtime, control-events, schema, validation
kv-canary: must have dtype , got	validation	error	kv-canary, dtype, validation
mm_preprocess_cache_size_mb must be non-negative	validation	error	multimodal, cache, server-args, validation
mm_process_config must be a dict, but got	validation	error	multimodal, json, server-args, type-error
no Cargo package under	exception	error	rust, discovery, module-not-found, metadata
No trace files found for profile_id	validation	error	profiling, trace-merge, chrome-trace, sglang
nvImageCodec could not decode the JPEG image	exception	error	nvjpeg, image-decoding, gpu, multimodal, sglang
Some parameters like	validation	critical	checkpoint-loading, mtp, random-init-guard
SSL CA certificates file not found	validation	error	ssl, tls, certificates, server-args, startup-validation
thinking content parts are only valid in assistant messages	validation	error	openai-api, thinking-part, role-validation
tokenspeed_mla backend requires kv-cache-dtype=fp8_e4m3, got	validation	error	sglang, tokenspeed, mla, kv-cache-dtype, config-validation
Unsupported time_compression_ratio	validation	error	python, value-error, model-config, hunyuan-vae, compression-ratio
appendExternalCorpusTokens called without…	exception	error	ngram, corpus-loading, api-misuse, state-machine
Block sparse tensors	validation	error	block-sparse, shape-mismatch, consistency
Comfy full_precision_matrix_mult does not support fused…	validation	error	quantization, nvfp4, comfy, fused-layers
component_attention_backends must use component=backend…	validation	error	config, cli, validation, attention-backend
Cosmos3 accepts either --image-path (I2V) or --video-path…	validation	error	cosmos3, i2v, v2v, mutually-exclusive, conditioning
Currently, only 4bits is supported on CPU with AMX.	validation	error	gptq, quantization, cpu, amx, unsupported-operation
Failed to load image_processor for	exception	error	multimodal, image-processor, dependencies
hidden_size must be divisible by num_heads	validation	error	stablelm, tensor-parallel, divisibility
kv-canary: offsets kernel bs must be in	validation	error	kv-canary, batch-size, bounds-check
kv-canary: scatter_req_token_ids offsets must be 1-D, got…	validation	error	kv-cache, shape-validation, tensor-rank
MiniMax-H3 quality="high" requires a resolved request plan	validation	error	minimax-h3, quality-high, request-plan, pipeline-order
Missing tensor payload for module(s)	validation	error	weights-update, payload, validation, multimodal
Q8KV8 sparse-prefill topk width must be a positive multiple…	validation	error	shape-validation, topk, sparse-attention
reference audio is empty	validation	error	minimax-h3, audio, empty-media
not supported. Choose between 'interleaved' and 'split'.	exception	critical	config-validation, rope, init-time, ltx2
SANA-WM Triton camera GDN backend unavailable	error_code	error	sana-wm, gdn, triton, camera-branch, shape-constraints
`sglang.bench_one_batch` is deprecated and will be removed…	console	warning	deprecation, benchmark, one-batch, future-warning
Tensor parallel size	validation	error	step3-vl, tensor-parallel, moe
The hpc_ops MoE runner backend does not support no_combine…	validation	error	sglang, moe, hpc-ops, no-combine, config-validation
The layout of k is not supported	exception	error	cuda, flash-attention, memory-layout, sm100, kv-cache
total_verify_tokens != sum(verify_lens_cpu)	validation	error	speculative-decoding, validation, invariant
VisionAttention(head_size=...) is deprecated; use…	console	warning	deprecation, sglang, vision, api-rename
action must have shape [T, D], got	validation	error	cosmos3, action-generation, tensor-shape, validation
bad compress_ratio	validation	error	deepseek, sparse-attention, indexing, config-validation
--dcp-comm-backend only affects the decode context-parallel…	validation	error	server-args, dcp, comm-backend, parallelism, sglang
--enable-svdquant cannot be combined with a GGUF…	validation	error	gguf, svdquant, nunchaku, config-conflict
Expected a string in the format	validation	error	registry, model-registration, lazy-import, validation
Failed to fit coefficients: insufficient rank	validation	error	pipeline-parallel, profiling, linear-algebra
flattened_bucket payload missing 'flattened_tensor' or…	validation	error	weights-update, flattened-bucket, missing-key, multimodal
Guidance scale must be positive, but got	validation	error	cfg, guidance-scale, input-validation, range-check
LogicalHostPool allocation must be page-aligned, got…	validation	error	sglang, memory-pool, allocation, page-alignment
LSE tensor must be Float32	validation	error	cuda, dtype, flash-attention, cutlass, lse
No tool calls but found tool output	exception	error	deepseek, tool-calls, validation
prompt event payload must be a string	validation	error	realtime, event-validation, prompt, websocket
QKV tensors must be on the same CUDA device	validation	error	device, multi-gpu, rope, hunyuan