private-gpt
private-gpt copied to clipboard
Embedding reading PGPT_MODE key
Hi all,
I'm installing privategpt 0.4.0. I've configured the setup with PGPT_MODE = openailike. Some of my settings are as follows: llm: mode: openailike max_new_tokens: 10000 context_window: 26000
embedding: mode: huggingface
huggingface: embedding_hf_model_name: ${PGPT_EMBEDDING_HF_MODEL_NAME:BAAI/bge-large-en-v1.5}
openai: api_base: http://172.20.0.2:5000/v1 api_key: EMPTY model: facebook/opt-125m prompt_style: "mistral"
But when trying to run the docker image I keep getting the error :
privategpt-20240307-private-gpt-1 | 17:26:15.713 [INFO ] private_gpt.settings.settings_loader - Starting application with profiles=['default', 'docker']
privategpt-20240307-private-gpt-1 | Traceback (most recent call last):
privategpt-20240307-private-gpt-1 | File "
I think in some part the embeddings model uses PGPT_KEY instead of the embedding->mode key. Or maybe I'm wrong with my settings?
thank you very much