batch mismatch when using openai retriever

[zsychina]

locomo/task_eval/rag_utils.py

Line 134 in 3eb6f2c

all_embeddings.append(torch.tensor(get_openai_embedding(inputs)))

should be

all_embeddings.append(torch.tensor(get_openai_embedding(inputs[i:(i+batch_size)])))

otherwise will cause assert error in

locomo/task_eval/gpt_utils.py

Line 99 in 3eb6f2c

assert embeddings.shape[0] == len(dialogs), "Lengths of embeddings and dialogs do not match"