audio-flamingo-3 / llava /model /deprecate_consolidate.py
SreyanG-NVIDIA's picture
Upload 225 files
174ae06 verified
# Copyright (c) 2025 NVIDIA CORPORATION.
# Licensed under the MIT license.
# Adapted from https://github.com/NVlabs/VILA/tree/main under the Apache 2.0 license.
# LICENSE is in incl_licenses directory.
# Copyright 2024 NVIDIA CORPORATION & AFFILIATES
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# SPDX-License-Identifier: Apache-2.0
"""
Usage:
python3 -m llava.model.consolidate --src ~/model_weights/llava-7b --dst ~/model_weights/llava-7b_consolidate
"""
import argparse
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from llava.model import *
from llava.model.utils import auto_upgrade
def consolidate_ckpt(src_path, dst_path):
print("Loading model")
auto_upgrade(src_path)
src_model = AutoModelForCausalLM.from_pretrained(src_path, torch_dtype=torch.float16, low_cpu_mem_usage=True)
src_tokenizer = AutoTokenizer.from_pretrained(src_path, use_fast=False)
src_model.save_pretrained(dst_path)
src_tokenizer.save_pretrained(dst_path)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--src", type=str, required=True)
parser.add_argument("--dst", type=str, required=True)
args = parser.parse_args()
consolidate_ckpt(args.src, args.dst)