Closed njalan closed 1 month ago
@hydai Can you please help to look at this?
Qwen is using the ChatML template format for prompts; you need to add the -p chatml
argument to your command. It should work out-of-the-box using the latest release of the GGML plugin.
Additionally, llama-chat.wasm
originates from LlamaEdge/chat; perhaps you should consider raising an issue here as a first step next time.
@dm4 Thanks for your help
In my testing qwen1_5-14b-chat-q5_k_m.gguf is working with llama.cpp but not work with wasmedge & llama-chat.wasm.
Here is my command: wasmedge --dir .:. --nn-preload default:GGML:CPU:qwen1_5-14b-chat-q5_k_m.gguf llama-chat.wasm -c 2048 -n 512 -g 0 there is nothing output