EVOKE
γICLR 2026 π₯γThis work introduces MMEVOKE benchmark to reveal challenges in knowledge injection and explores potential solutions.
νμΌ νμκΈ°
μ΅μ’ λ²μ λ€μ΄λ‘λ (.zip)- all_type_score.py
- eval_acc_f1.py
- tools.py
- vqa.py
- data_construction.png
- data_display.png
- motivation.jpg
- devcontainer.env
- devcontainer.json
- Dockerfile
- postCreateCommand.sh
- 1-usage.yaml
- 2-feature-request.yaml
- 3-question.yaml
- 4-discussion.yaml
- alpaca.png
- bard.jpg
- chatgpt.svg
- llama.jpg
- swords_FILL0_wght300_GRAD0_opsz48.svg
- vicuna.jpeg
- index.html
- script.js
- styles.css
- eval_gpt_review.py
- eval_gpt_review_bench.py
- eval_gpt_review_visual.py
- eval_pope.py
- eval_science_qa.py
- eval_science_qa_gpt4.py
- eval_science_qa_gpt4_requery.py
- eval_textvqa.py
- generate_webpage_data_from_table.py
- m4c_evaluator.py
- model_qa.py
- model_vqa.py
- model_vqa_loader.py
- model_vqa_mmbench.py
- model_vqa_science.py
- qa_baseline_gpt35.py
- run_llava.py
- summarize_gpt_review.py
- llava_llama.py
- llava_mistral.py
- llava_mpt.py
- builder.py
- clip_encoder.py
- builder.py
- __init__.py
- apply_delta.py
- builder.py
- consolidate.py
- llava_arch.py
- make_delta.py
- utils.py
- extreme_ironing.jpg
- waterview.jpg
- __init__.py
- cli.py
- controller.py
- gradio_web_server.py
- model_worker.py
- register_worker.py
- sglang_worker.py
- test_message.py
- llama_flash_attn_monkey_patch.py
- llama_xformers_attn_monkey_patch.py
- llava_trainer.py
- train.py
- train_mem.py
- train_xformers.py
- __init__.py
- constants.py
- conversation.py
- mm_utils.py
- utils.py
- gqa.sh
- llavabench.sh
- mmbench.sh
- mmbench_cn.sh
- mme.sh
- mmvet.sh
- pope.sh
- qbench.sh
- qbench_zh.sh
- seed.sh
- sqa.sh
- textvqa.sh
- vizwiz.sh
- vqav2.sh
- finetune.sh
- finetune_lora.sh
- finetune_task.sh
- finetune_task_lora.sh
- pretrain.sh
- convert_gqa_for_eval.py
- convert_mmbench_for_submission.py
- convert_mmvet_for_eval.py
- convert_seed_for_submission.py
- convert_sqa_to_llava.py
- convert_sqa_to_llava_base_prompt.py
- convert_vizwiz_for_submission.py
- convert_vqav2_for_submission.py
- extract_mm_projector.py
- finetune.sh
- finetune_full_schedule.sh
- finetune_lora.sh
- finetune_qlora.sh
- finetune_sqa.sh
- merge_lora_weights.py
- pretrain.sh
- pretrain_xformers.sh
- sqa_eval_batch.sh
- sqa_eval_gather.sh
- upload_pypi.sh
- .dockerignore
- .editorconfig
- .gitattributes
- .gitignore
- cog.yaml
- LICENSE
- mm_rag_llava_inference.py
- predict.py
- pyproject.toml
- README.md
- bug_report.yaml
- config.yaml
- feature_request.yaml
- Beijing.jpeg
- Beijing_Small.jpeg
- Chongqing.jpeg
- Chongqing_Small.jpeg
- Hospital.jpg
- Hospital_Small.jpg
- Menu.jpeg
- Rebecca_(1939_poster).jpeg
- Rebecca_(1939_poster)_Small.jpeg
- Shanghai.jpg
- Shanghai_Output.jpg
- Shanghai_Output_Small.jpeg
- Shanghai_Small.jpeg
- TUTORIAL.ipynb
- apple.jpeg
- apple_r.jpeg
- demo.jpeg
- demo_highfive.jpg
- demo_spotting_caption.jpg
- demo_vl.gif
- logo.jpg
- qwenvl.jpeg
- radar.png
- radar_qwenvlplus.jpg
- touchstone_datasets.jpg
- touchstone_eval.png
- touchstone_logo.png
- wechat.png
- evaluate_multiple_choice_mmbench.py
- MMBENCH.md
- mmbench_converter_dev.py
- mmbench_converter_test.py
- mmbench_evaluation.py
- mmbench_evaluation_tricky.py
- mmbench_predict_to_submission.py
- cognition.jpg
- eval.py
- EVAL_MME.md
- get_images.py
- perception.jpg
- eval.py
- EVAL_SEED.md
- leaderboard.jpg
- trans.py
- eval_bash.sh
- evaluate_caption.py
- evaluate_grounding.py
- evaluate_multiple_choice.py
- evaluate_vqa.py
- EVALUATION.md
- infographicsvqa_eval.py
- test.py
- vqa.py
- vqa_eval.py
- ds_config_zero2.json
- ds_config_zero3.json
- finetune_ds.sh
- finetune_ds_bash_1.sh
- finetune_lora_ds.sh
- finetune_lora_ds_bash_1.sh
- finetune_lora_single_gpu.sh
- finetune_qlora_ds.sh
- finetune_qlora_single_gpu.sh
- README.md
- README_CN.md
- README_JA.md
- README_KO.md
- .gitignore
- BUILD.md
- Dockerfile.qwendemo
- Dockerfile.qwenint4openai
- Dockerfile.qwenopenai
- FAQ.md
- FAQ_ja.md
- FAQ_ko.md
- FAQ_zh.md
- finetune.py
- LICENSE
- NOTICE
- openai_api.py
- pip.sh
- README.md
- README_CN.md
- README_JA.md
- README_KO.md
- requirements.txt
- requirements_openai_api.txt
- requirements_web_demo.txt
- TUTORIAL.md
- TUTORIAL_ja.md
- TUTORIAL_ko.md
- TUTORIAL_zh.md
- web_demo_mm.py
- retrieval_data_with_golden_context.jsonl
- retrieval_data_with_top_k_image_only.jsonl
- retrieval_data_with_top_k_text_only.jsonl
- retrieval_data_with_top_k_UniIR.jsonl
- mbeir_cirr_task7_cand_pool.jsonl
- mbeir_new_self_cand_pool.jsonl
- query_instructions.tsv
- mbeir_cirr_task7_test_qrels.txt
- mbeir_new_self_test_qrels.txt
- mbeir_cirr_task7_test.jsonl
- mbeir_new_self_test.jsonl
- .gitattributes
- .gitignore
- README.md
- uniir_teaser.jpg
- mbeir_new_self_cand_pool_embed.npy
- mbeir_new_self_cand_pool_ids.npy
- mbeir_union_cand_pool_embed.npy
- mbeir_union_cand_pool_ids.npy
- mbeir_new_self_test_embed.npy
- mbeir_new_self_test_ids.npy
- mbeir_new_self_cand_pool.index
- mbeir_union_cand_pool.index
- mbeir_new_self_union_pool_test_k10_run.txt
- mbeir_new_self_union_pool_test_k10_run_new_train_0_10.txt
- mbeir_new_self_union_pool_test_k10_run_new_train_0_2.txt
- mbeir_new_self_union_pool_test_k10_run_new_train_0_4.txt
- mbeir_new_self_union_pool_test_k10_run_new_train_0_6.txt
- mbeir_new_self_union_pool_test_k10_run_new_train_0_8.txt
- config_updater.py
- dist_utils.py
- faiss_env.yml
- interactive_retriever.py
- mbeir_analyst.py
- mbeir_embedder.py
- mbeir_retriever.py
- res.txt
- run.py
- utils.py
- __init__.py
- cirr_data_preprocessor.py
- edis_data_preprocessor.py
- fashion200k_data_preprocessor.py
- fashioniq_data_preprocessor.py
- infoseek_data_preprocessor.py
- mscoco_data_preprocessor.py
- nights_data_preprocessor.py
- oven_data_preprocessor.py
- utils.py
- visualnews_data_preprocessor.py
- webqa_data_preprocessor.py
- __init__.py
- mbeir_data_generator.py
- mbeir_data_utils.py
- mbeir_dataset.py
- bert_config.json
- med_config.json
- __init__.py
- blip_transform.py
- randaugment.py
- __init__.py
- blip.py
- med.py
- vit.py
- embed.yaml
- index.yaml
- retrieval.yaml
- run_eval_pipeline_inbatch.sh
- inbatch.yaml
- run_inbatch.sh
- embed.yaml
- index.yaml
- retrieval.yaml
- run_eval_pipeline_inbatch.sh
- embed.yaml
- index.yaml
- retrieval.yaml
- run_eval_pipeline_unirag.sh
- inbatch.yaml
- run_inbatch.sh
- __init__.py
- blip_ff.py
- embed.yaml
- index.yaml
- retrieval.yaml
- run_eval_pipeline_inbatch.sh
- inbatch.yaml
- run_inbatch.sh
- embed.yaml
- index.yaml
- retrieval.yaml
- run_eval_pipeline_inbatch.sh
- inbatch.yaml
- run_inbatch.sh
- __init__.py
- blip_sf.py
- __init__.py
- engine.py
- env.yml
- train.py
- utils.py
- embed.yaml
- index.yaml
- retrieval.yaml
- run_eval_pipeline_inbatch.sh
- inbatch.yaml
- run_inbatch.sh
- embed.yaml
- index.yaml
- retrieval.yaml
- run_eval_pipeline_inbatch.sh
- inbatch.yaml
- run_inbatch.sh
- __init__.py
- clip_ff.py
- train.py
- embed.yaml
- index.yaml
- retrieval.yaml
- run_eval_pipeline_inbatch.sh
- run_eval_pipeline_inbatch_no_instruct.sh
- inbatch.yaml
- run_inbatch.sh
- run_inbatch_no_instruct.sh
- embed.yaml
- index.yaml
- retrieval.yaml
- run_eval_pipeline_inbatch.sh
- run_eval_pipeline_inbatch_no_instruct.sh
- embed.yaml
- index.yaml
- retrieval.yaml
- retrieval_ori.yaml
- run_eval_pipeline_unirag.sh
- inbatch.yaml
- run_inbatch.sh
- run_inbatch_no_instruct.sh
- __init__.py
- clip_sf.py
- train.py
- __init__.py
- engine.py
- env.yml
- utils.py
- __init__.py
- uniir_env.yml
- __init__.py
- .gitignore
- data_self.pkl
- ii_qrel.txt
- LICENSE
- mbeir_oven_task8_cand_pool_short.jsonl
- mbeir_oven_task8_test_qrels.txt
- mbeir_oven_task8_test_short.jsonl
- ori_format_new_self.py
- pool_size_format_new_self.py
- query_instructions.tsv
- README.md
- retrieval_image_only.py
- retrieval_text_only.py
- README.md
# μ€μΉ κ°μ΄λ
1. μ½λ λ΄λ €λ°κΈ°
git clone https://github.com/EVOKE-LMM/EVOKE
κΉνλΈμμ νλ‘μ νΈ μ½λ μ 체λ₯Ό λ΄ μ»΄ν¨ν°λ‘ λ΄λ €λ°μ΅λλ€.
cd EVOKE
λ°©κΈ λ΄λ €λ°μ νλ‘μ νΈ ν΄λ μμΌλ‘ μ΄λν©λλ€.
2. Docker
μ¬μ μΆμ²μ¬μ μ€λΉλ¬Ό
- Git GitHubμμ νλ‘μ νΈ μ½λλ₯Ό λ΄λ €λ°μΌλ €λ©΄ νμν©λλ€.
- Docker Desktop 컨ν μ΄λλ₯Ό λΉλνκ³ μ€ννλ €λ©΄ νμν©λλ€. μ€μΉ ν μ€νν΄μ λ°±κ·ΈλΌμ΄λμ μΌλμΈμ.
docker build -f Qwen-VL/Dockerfile.qwendemo -t evoke .
Dockerfileμ κΈ°λ°μΌλ‘ μ€ν κ°λ₯ν μ΄λ―Έμ§λ₯Ό λΉλν©λλ€.
docker run -p 8080:80 evoke
λΉλλ μ΄λ―Έμ§λ₯Ό μ€μ 컨ν μ΄λλ‘ μ€νν©λλ€.
ν°λ―Έλμ docker compose ps λ₯Ό μ
λ ₯ν΄ μ»¨ν
μ΄λλ€μ΄ Up μνμΈμ§ νμΈνμΈμ. READMEμ ν¬νΈ λ²νΈκ° μ νμλ€λ©΄ λΈλΌμ°μ μμ http://localhost:ν¬νΈλ²νΈ λ‘ μ μν΄λ³΄μΈμ.
3. Python
μ¬μμ¬μ μ€λΉλ¬Ό
pip install -r Qwen-VL/requirements.txt
requirements.txt λ±μ λͺ μλ νμ΄μ¬ λΌμ΄λΈλ¬λ¦¬λ₯Ό μ€μΉν©λλ€.
jupyter notebook
λΈλΌμ°μ μμ λ ΈνΈλΆ(.ipynb) νμΌλ€μ μ΄λνκ³ μ€νν μ μλ Jupyter νλ©΄μ μΌλλ€.
μλ¬ λ©μμ§ μμ΄ μ€νλκ³ ν°λ―Έλμ μλ΄ λ¬Έκ΅¬κ° μΆλ ₯λλ©΄ μ μμ
λλ€.
// repository documentation
Was this content helpful?
(0 ratings)