eval/eval_esm.py

import argparse
import os
import random
import sys
import time
import tqdm
sys.path.insert(0, "..")

import numpy as np
import torch
import torch.backends.cudnn as cudnn

from minigpt4.common.config import Config
from minigpt4.common.dist_utils import get_rank
from minigpt4.common.registry import registry
from minigpt4.conversation.conversation_esm import Chat, CONV_VISION

# imports modules for registration
from minigpt4.datasets.builders import *
from minigpt4.models import *
from minigpt4.processors import *
from minigpt4.runners import *
from minigpt4.tasks import *
import sys

import esm
import json

DATASET_SPEC = "/home/ubuntu/proteinchat/dataset.json"

ANN_PATH = "/home/ubuntu/proteinchat/data/qa_all.json"
PDB_PATH = "/home/ubuntu/pt"
SEQ_PATH = "/home/ubuntu/seq"

OUTPUT_SAVE_PATH = "/home/ubuntu/proteinchat/eval/results/outputs"
annotation = open(ANN_PATH, "r")
annotation = json.load(annotation)

dataset = open(DATASET_SPEC, "r")
dataset = json.load(dataset)
all_prots = dataset["test"] 

def parse_args():
    parser = argparse.ArgumentParser(description="Demo")
    parser.add_argument("--cfg-path", required=True, help="path to configuration file.")
    parser.add_argument("--gpu-id", type=int, default=0, help="specify the gpu to load the model.")
    parser.add_argument("--model", type=str, required=True, help="specify the model to load the model.")
    parser.add_argument(
        "--options",
        nargs="+",
        help="override some settings in the used config, the key-value pair "
        "in xxx=yyy format will be merged into config file (deprecate), "
        "change to --cfg-options instead.",
    )
    args = parser.parse_args()
    return args


def setup_seeds(config):
    seed = config.run_cfg.seed + get_rank()

    random.seed(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)

    cudnn.benchmark = False
    cudnn.deterministic = True

print('Initializing Chat')
args = parse_args()
cfg = Config(args)

model_config = cfg.model_cfg
model_config.device_8bit = args.gpu_id
model_cls = registry.get_model_class(model_config.arch)
model = model_cls.from_config(model_config).to('cuda:{}'.format(args.gpu_id))

vis_processor_cfg = cfg.datasets_cfg.cc_sbu_align.vis_processor.train
vis_processor = registry.get_processor_class(vis_processor_cfg.name).from_config(vis_processor_cfg)
chat = Chat(model, vis_processor, device='cuda:{}'.format(args.gpu_id)) 
print('Initialization Finished')

raw_output = {}
score_output = {}

START_SAMPLES = 0
# END_SAMPLES = 8806
END_SAMPLES = 160

all_prots = all_prots[START_SAMPLES : END_SAMPLES]

for prot in tqdm.tqdm(all_prots):
    curr_prot_ann = annotation[prot]
    pdb_path = os.path.join(PDB_PATH, f"{prot}.pt")
    seq_path = os.path.join(SEQ_PATH, f"{prot}.pt") 
    seq_embedding = torch.load(seq_path, map_location=torch.device('cpu'))
    sample_seq = seq_embedding.to('cuda:{}'.format(args.gpu_id))

    if (seq_embedding.shape[1] > 384):
        continue
    raw_output[prot] = []
    pdb_embedding = torch.load(pdb_path, map_location=torch.device('cpu'))
    sample_pdb = pdb_embedding.to('cuda:{}'.format(args.gpu_id))
           
    
    for ann in curr_prot_ann:
        d = {}
        d["Q"] = ann["Q"]
        chat_state = CONV_VISION.copy()
        img_list = []
        llm_message = chat.upload_protein(sample_pdb, sample_seq, chat_state, img_list)
        img_list = [mat.half() for mat in img_list]
        chat.ask(ann["Q"], chat_state)
        ans = chat.answer(conv=chat_state,
                                img_list=img_list,
                                num_beams=1,
                                temperature=0.7,
                                max_new_tokens=384,
                                max_length=2048)[0]
        d["A"] = ans
        raw_output[prot].append(d)
        
with open(os.path.join(OUTPUT_SAVE_PATH, f"{args.model}_eval_output.json"), 'w') as fp:
    json.dump(raw_output, fp, indent=4)