mirror of
https://github.com/microsoft/semantic-kernel.git
synced 2026-10-12 02:19:06 +00:00
Karol ŻakandAbby Harrison
601c9de124
### Motivation and Context <!-- Thank you for your contribution to the semantic-kernel repo! Please help reviewers and future users, providing the following information: 1. Why is this change required? 2. What problem does it solve? 3. What scenario does it contribute to? 4. If it fixes an open issue, please link to the issue here. --> I was trying to use this library to load and use [HF Falcon model](https://huggingface.co/tiiuae/falcon-7b-instruct). Unfortunately in order to make it work I needed to pass in some additional arguments such as `torch_dtype`, `cache_dir` or `tokenizer`. It was not possible with current implementation so I decided to add `model_kwargs` and `pipeline_kwargs` to the `HuggingFaceTextCompletion` class init function so now, with the changes I introduced I can successfully load any `falcon` models by using this code: ```python import semantic_kernel as sk import semantic_kernel.connectors.ai.hugging_face as sk_hf import torch from transformers import AutoTokenizer cache_dir = "/mnt/tmp" model_name = "tiiuae/falcon-7b-instruct" tokenizer = AutoTokenizer.from_pretrained( pretrained_model_name_or_path=model_name, cache_dir=cache_dir, trust_remote_code=True) hf_model = sk_hf.HuggingFaceTextCompletion( model_name, task="text-generation", model_kwargs={ "cache_dir": cache_dir, "torch_dtype": torch.bfloat16 }, pipeline_kwargs={ "tokenizer": tokenizer, "trust_remote_code": True, "device": 0, } ) ``` ### Description <!-- Describe your changes, the overall approach, the underlying design. These notes will help understanding how your code works. Thanks! --> The changes I introduced add `model_kwargs` and `pipeline_kwargs` to `HuggingFaceTextCompletion` class so that these arguments are then passed to the underlying `transformers.pipeline()` function and thus allow more customization options. ### Contribution Checklist <!-- Before submitting this PR, please make sure: --> - [x] The code builds clean without any errors or warnings - [x] The PR follows the [SK Contribution Guidelines](https://github.com/microsoft/semantic-kernel/blob/main/CONTRIBUTING.md) and the [pre-submission formatting script](https://github.com/microsoft/semantic-kernel/blob/main/CONTRIBUTING.md#development-scripts) raises no violations - [x] All unit tests pass, and I have added new tests where possible - [x] I didn't break anyone 😄 --------- Co-authored-by: Abby Harrison <[email protected]>