Closed leonjza closed 5 months ago
Hey! Thanks a lot!
Could you please also add your pattern to the table in the readme and (if possible) upload a small test file to tests/patterns/test_data
so the CI can run against it and make sure it doesn't break in the future?
Added a test file. It's a LLama2 model with the tensor data truncated (which is just a blob where a structure has an offset to read data from and not needed for parsing). The newline / "unrelated" changes in the README is a markdownlint run. Let me know if you'd prefer I remove those.
Start 36: Patterns/gguf.hexpat/gguf.hexpat.gguf
36/145 Test #36: Patterns/gguf.hexpat/gguf.hexpat.gguf .......................... Passed 9.04 sec
Thank you!
Add's a pattern for GGUF formatted models. Ref: https://github.com/ggerganov/ggml/blob/master/docs/gguf.md
A sample file can be grabbed from hugging face here: https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v0.6/blob/main/ggml-model-q4_0.gguf