diff --git a/README.md b/README.md index 44dd2b8..4adaec9 100644 --- a/README.md +++ b/README.md @@ -81,6 +81,7 @@ See: [OpenVLM Leaderboard](https://huggingface.co/spaces/opencompass/open_vlm_le Version 0.14.1 +- new model support: idefics2-8b-chatty, idefics2-8b-chatty-AWQ (it worked already, no code change) - new model support: XComposer2-VL-1.8B (it worked already, no code change) Version: 0.14.0 diff --git a/model_conf_tests.json b/model_conf_tests.json index 1d23495..d76322b 100644 --- a/model_conf_tests.json +++ b/model_conf_tests.json @@ -1,6 +1,7 @@ [ ["vikhyatk/moondream2", "--use-flash-attn"], ["OpenGVLab/InternVL-Chat-V1-5", "--device-map", "cuda:0"], + ["HuggingFaceM4/idefics2-8b-chatty", "--use-flash-attn", "--device-map", "cuda:0"], ["HuggingFaceM4/idefics2-8b", "--use-flash-attn", "--device-map", "cuda:0"], ["qihoo360/360VL-8B", "--use-flash-attn"], ["qnguyen3/nanoLLaVA", "--use-flash-attn", "--device-map", "cuda:0"], @@ -33,6 +34,7 @@ ["OpenGVLab/InternVL-Chat-V1-5-Int8", "--device-map", "cuda:0"], ["OpenGVLab/InternVL-Chat-V1-5", "--load-in-4bit", "--device-map", "cuda:0"], + ["HuggingFaceM4/idefics2-8b-chatty-AWQ", "--use-flash-attn", "--device-map", "cuda:0"], ["HuggingFaceM4/idefics2-8b-AWQ", "--use-flash-attn", "--device-map", "cuda:0"], ["qihoo360/360VL-8B", "--use-flash-attn", "--load-in-4bit"], ["qihoo360/360VL-70B", "--use-flash-attn", "--load-in-4bit"], diff --git a/vision.sample.env b/vision.sample.env index 4941c78..e9962a7 100644 --- a/vision.sample.env +++ b/vision.sample.env @@ -4,6 +4,7 @@ HF_HOME=hf_home #CUDA_VISIBLE_DEVICES=1,0 #CLI_COMMAND="python vision.py -m vikhyatk/moondream2 --use-flash-attn" # test pass✅, time: 4.1s, mem: 4.6GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m OpenGVLab/InternVL-Chat-V1-5 --device-map cuda:0" # test pass✅, time: 13.4s, mem: 52.0GB, 8/8 tests passed. +#CLI_COMMAND="python vision.py -m HuggingFaceM4/idefics2-8b-chatty --use-flash-attn --device-map cuda:0" # test pass✅, time: 8.6s, mem: 22.5GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m HuggingFaceM4/idefics2-8b --use-flash-attn --device-map cuda:0" # test pass✅, time: 6.9s, mem: 22.4GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m qihoo360/360VL-8B --use-flash-attn" # test pass✅, time: 5.7s, mem: 17.4GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m qnguyen3/nanoLLaVA --use-flash-attn --device-map cuda:0" # test pass✅, time: 7.6s, mem: 8.0GB, 8/8 tests passed. @@ -34,6 +35,7 @@ HF_HOME=hf_home #CLI_COMMAND="python vision.py -m YanweiLi/MGM-2B --use-flash-attn" # test pass✅, time: 4.4s, mem: 8.3GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m OpenGVLab/InternVL-Chat-V1-5-Int8 --device-map cuda:0" # test pass✅, time: 26.9s, mem: 32.0GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m OpenGVLab/InternVL-Chat-V1-5 --load-in-4bit --device-map cuda:0" # test fail❌, time: 17.2s, mem: 18.2GB, 2/8 tests passed. +#CLI_COMMAND="python vision.py -m HuggingFaceM4/idefics2-8b-chatty-AWQ --use-flash-attn --device-map cuda:0" # test pass✅, time: 9.9s, mem: 12.7GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m HuggingFaceM4/idefics2-8b-AWQ --use-flash-attn --device-map cuda:0" # test pass✅, time: 8.0s, mem: 12.6GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m qihoo360/360VL-8B --use-flash-attn --load-in-4bit" # test pass✅, time: 8.5s, mem: 7.8GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m qnguyen3/nanoLLaVA --use-flash-attn --load-in-4bit --device-map cuda:0" # test pass✅, time: 10.5s, mem: 7.6GB, 8/8 tests passed.