Garbish output
Hi! On my side both 4-bit (Q4_0) and 5-bit (Q5_K_M) versions return garbage.
Tried from GPT4All UI and LamaEdge chat with recommended parameters. Any comments\recommendation?
wasmedge --dir .:. --nn-preload default:GGML:AUTO:Llama-3.1-Nemotron-70B-Reward-HF-Q5_K_M.gguf
llama-chat.wasm
--prompt-template llama-3-chat
--ctx-size 128000
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