CUDA_ERROR_INSUFFICIENT_DRIVER warning CUDA driver version mismatch
Installed CUDA toolkit version is newer than what the installed driver supports.
View detailsError directory
Common failures during local inference, grouped by the module that throws them.
CUDA
CUDA_OUT_OF_MEMORY critical CUDA error: out of memory
The model, context length, or batch size requested more VRAM than the GPU has available.
View detailsCUDA_ERROR_OUT_OF_MEMORY critical CUDA out of memory
CUDA_ERROR_OUT_OF_MEMORY means an allocation exceeded free VRAM. Check what else holds VRAM, then lower the context size or use a smaller quantization.
View detailsOllama
OLLAMA_CONNECTION_REFUSED info Error: could not connect to Ollama, is it running?
The Ollama server isn't running, or a client is pointed at the wrong host/port.
View detailsOLLAMA_MODEL_NOT_FOUND info Ollama model not found locally
The model tag hasn't been pulled yet, or the tag name doesn't match the registry exactly.
View detailsvLLM
VLLM_MAX_SEQ_LEN_EXCEEDED warning ValueError: prompt exceeds the model's maximum context length
The request's prompt + max_tokens exceeds max_model_len, often after that value was manually raised past what KV cache memory supports.
View detailsVLLM_MAX_SEQ_LEN_EXCEEDED warning ValueError: prompt exceeds the model's maximum context length
The request's prompt + max_tokens exceeds max_model_len, often after that value was manually raised past what KV cache memory supports.
View detailsVLLM_TENSOR_PARALLEL_CRASH critical vLLM tensor-parallel crash on startup
Tensor-parallel size doesn't evenly divide the model's attention heads.
View details