Failed to build the environment.

[zhuyjan]

When using pip install -r requirements.txt, the following error occurs:

ERROR: Invalid requirement: '_libgcc_mutex=0.1=conda_forge': Expected package name at the start of dependency specifier
    _libgcc_mutex=0.1=conda_forge
    ^ (from line 4 of requirements.txt)
Hint: = is not a valid operator. Did you mean == ?

It seems that the format of this requirements.txt file is unusual. Could you please advise on the correct way to use this configuration file in order to properly set up the environment?

---

Would you like me to also suggest a short clarification in your issue that this file may have been generated from conda (not pip), so the maintainers immediately understand the root cause?

[VAN-QIAN]

Hi @zhuyjan , I am not the author but I have tried to build the environment by myself. Hope this can help.

1. Create the conda environment and install torch

conda create -n reflectiva python=3.8.16 -y
conda activate reflectiva
pip install "torch==2.3.0+cu118" "torchvision==0.18.0+cu118" -f https://download.pytorch.org/whl/cu118/torch_stable.html

2. Use the file uploaded requirements-pip.txt to install packages via pip.

pip install -r requirements-pip.txt

3. On my machine I can run the evaluation successfully via the following command after downloading the data files following all the instruction. However, my issue is that the evaluation speed is a bit slow as each iteration will takes around 30 seconds on one Nvidia A6000 gpu. I am not sure if this is caused by the relatively out-of-date processor core of my gpu and I would appreciate it if anyone or the authors can join the discussion.

export PYTHONPATH=.
model=aimagelab/ReflectiVA
export TOKENIZER_PATH=$model
export PART=0
export TOTAL_PART=0
nohup python rag_evaluation/encyclopedic/release_retrieval.py \
--model_path $model \
--model_name llava_llama_3.1 \
--data_path ~/ReflectiVA/data_evqa/test.json \
--index_path ~/ReflectiVA/evqa_EVA_image \
--index_path_json ~/ReflectiVA/evqa_EVA_image \
--entity_k 5 \
--use_eva_to_retrieve \
--retriever_path BAAI/EVA-CLIP-8B \
--retriever_processor_path openai/clip-vit-large-patch14 \
--kb_wikipedia_path ~/ReflectiVA/data_evqa/encyclopedic_kb_wiki.json \
--short_prompt > eval_evqa_full.log 2>&1 &