KeyLLM fails when no GPU is available

[lincis]

KeyLLM fails when no GPU is available:
`

RuntimeError Traceback (most recent call last)
Cell In[14], line 13
10 kw_model = KeyBERT(llm = llm)
12 # Extract keywords
---> 13 keywords = kw_model.extract_keywords(sparks.query('id == "4d430272-7668-4353-b113-d1ebdda308a4"')['full_text'].values[0])

File ~/Dokumenti/Python_projects/DS-prototypes/Spark summarization/.venv/lib/python3.11/site-packages/keybert/_model.py:259, in KeyBERT.extract_keywords(self, docs, candidates, keyphrase_ngram_range, stop_words, top_n, min_df, use_maxsum, use_mmr, diversity, nr_candidates, vectorizer, highlight, seed_keywords, doc_embeddings, word_embeddings, threshold)
257 if self.llm is not None:
258 import torch
--> 259 doc_embeddings = torch.from_numpy(doc_embeddings).float().to("cuda")
260 if isinstance(all_keywords[0], tuple):
261 candidate_keywords = [[keyword[0] for keyword in all_keywords]]

File ~/Dokumenti/Python_projects/DS-prototypes/Spark summarization/.venv/lib/python3.11/site-packages/torch/cuda/init.py:298, in _lazy_init()
296 if "CUDA_MODULE_LOADING" not in os.environ:
297 os.environ["CUDA_MODULE_LOADING"] = "LAZY"
--> 298 torch._C._cuda_init()
299 # Some of the queued calls may reentrantly call _lazy_init();
300 # we need to just return without initializing in that case.
301 # However, we must not let any other threads in!
302 _tls.is_initializing = True

RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from https://www.nvidia.com/Download/index.aspx
`

Seems that the issue is here:

KeyBERT/keybert/_model.py

Line 259 in 6638d44

doc_embeddings = torch.from_numpy(doc_embeddings).float().to("cuda")

If I remove .to("cuda") everything works like a charm.

[MaartenGr]

Hmmm, not quite sure what the best solution would be here. Perhaps an automatic detection of the system and using that instead. I imagine mac users would run into similar issues.