| # Copyright (c) 2022, salesforce.com, inc. | |
| # All rights reserved. | |
| # SPDX-License-Identifier: BSD-3-Clause | |
| # For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause | |
| # Bleu_1: 0.832 | |
| # Bleu_2: 0.691 | |
| # Bleu_3: 0.556 | |
| # Bleu_4: 0.438 | |
| # METEOR: 0.317 | |
| # ROUGE_L: 0.620 | |
| # CIDEr: 1.461 | |
| # SPICE: 0.252 | |
| model: | |
| arch: instruct_vicuna7b # blip2 blip2_opt instruct_vicuna7b | |
| model_type: vicuna7b # pretrain pretrain_opt2.7b vicuna7b | |
| use_grad_checkpoint: False | |
| pretrained: ./output/BLIP2/train_vicua/20250320054/checkpoint_best.pth #/root/autodl-tmp/Blip2CC/output/BLIP2/train_vicua/20250319222/checkpoint_best.pth #/root/autodl-tmp/Blip2CC/output/BLIP2/train_vicua/20250206211/checkpoint_0.pth #/root/autodl-tmp/Blip2CC/output/BLIP2/train_vicua/20250206211/checkpoint_0.pth | |
| datasets: | |
| changechat_105k: # ChangeChat-105k change captioning dataset | |
| vis_processor: | |
| eval: | |
| name: "blip_image_eval" | |
| image_size: 224 | |
| text_processor: | |
| eval: | |
| name: "blip_caption" | |
| # build_info: | |
| # images: | |
| # storage: '/export/share/datasets/vision/coco/images/' | |
| run: | |
| task: captioning | |
| # optimizer | |
| batch_size_train: 4 | |
| batch_size_eval: 4 | |
| num_workers: 0 | |
| max_len: 30 | |
| min_len: | |
| num_beams: 1 | |
| prompt: 'Please briefly describe the changes in these two images.' | |
| # Please briefly describe the changes in these two images.// prompt在dataset的text_input那里修改就好 | |
| seed: 42 | |
| output_dir: "output/BLIP2/Caption_coco_opt2.7b" | |
| evaluate: True | |
| test_splits: ["test"] | |
| device: "cuda" | |
| world_size: 1 | |
| dist_url: "env://" | |
| distributed: True | |