--quantized-type is not user friendly

[danielzgtg]

Following error messages to specify mandatory parameters, I got to this point:

$ ./quantize --quantized-type F16 --model-path ~/Kokoro_espeak.gguf --quantized-model-path ./out.gguf
ERROR: 0 is not a valid quantization type.

Turns out --quantize-type is an int_param‽

TTS.cpp/examples/quantize/quantize.cpp

Line 33 in 4bf8c80

enum ggml_type qtype = static_cast<ggml_type>(*args.get_int_param("--quantized-type"));

I should be able to use human-readable names like I do with llama.cpp

./llama-quantize ./models/mymodel/ggml-model-f16.gguf ./models/mymodel/ggml-model-Q4_K_M.gguf Q4_K_M

The following human-friendly names should be supported, corresponding to those in https://huggingface.co/mmwillet2/Kokoro_GGUF/tree/main :

• F16
• Q4
• Q5
• Q8

[mmwillet]

Yeah using the enums is unideal (especially when we are only really supporting k-quant zero types and floating point types). I'll put in a small patch tonight.