Quantization not working for models with BPE tokenizer

[sanchitvj]

Meta Llama-3 cannot be quantized because it uses BPE tokenizer.

[sanchitvj]

subprocess.CalledProcessError: Command '['python3', 'convert.py', '/home/ubuntu/volume_2k/Capstone_5/src/grag/quantize/test_data/llama.cpp/models/Meta-Llama-3-8B-Instruct/']' returned non-zero exit status 1.
Encountered the above error while quantizing Llama-3-8B because it uses a BPE tokenizer. For quantizing that you have to use below command line:
python convert.py models/mymodel/ --vocab-type bpe
more here

• Add support for such models.
• Better error handling for models with different quantization mechanisms.
• Validate if LlamaCpp supports quantization for given model.

[sanchitvj]

This issue is solved by the #118 PR. Now LLMs both with and without BPE tokenizer are supported.