diff --git a/README.md b/README.md index 3816e0e..9ac8452 100644 --- a/README.md +++ b/README.md @@ -21,6 +21,7 @@ An OpenAI API compatible vision server, it functions like `gpt-4-vision-preview` - - [X] [XComposer2-7b-4bit](https://huggingface.co/internlm/internlm-xcomposer2-7b-4bit) (not recommended) - - [X] [XComposer2-VL](https://huggingface.co/internlm/internlm-xcomposer2-vl-7b) [pretrain] (wont gpu split) - - [X] [XComposer2-VL-4bit](https://huggingface.co/internlm/internlm-xcomposer2-vl-7b-4bit) +- - [X] [XComposer2-VL-1.8B](https://huggingface.co/internlm/internlm-xcomposer2-vl-1_8b) - [X] [HuggingFaceM4/idefics2](https://huggingface.co/HuggingFaceM4) - - [X] [idefics2-8b](https://huggingface.co/HuggingFaceM4/idefics2-8b) (main docker only, wont gpu split) - - [X] [idefics2-8b-AWQ](https://huggingface.co/HuggingFaceM4/idefics2-8b-AWQ) (main docker only, wont gpu split) @@ -78,6 +79,10 @@ See: [OpenVLM Leaderboard](https://huggingface.co/spaces/opencompass/open_vlm_le ## Recent updates +Version 0.14.1 + +- new model support: XComposer2-VL-1.8B (it worked already, no code change) + Version: 0.14.0 - docker-compose.yml: Assume the runtime supports the device (ie. nvidia) diff --git a/model_conf_tests.alt.json b/model_conf_tests.alt.json index 794457f..1c798f7 100644 --- a/model_conf_tests.alt.json +++ b/model_conf_tests.alt.json @@ -21,6 +21,7 @@ ["internlm/internlm-xcomposer2-4khd-7b", "--use-flash-attn", "--device-map", "cuda:0"], ["internlm/internlm-xcomposer2-7b", "--use-flash-attn", "--device-map", "cuda:0"], ["internlm/internlm-xcomposer2-vl-7b", "--use-flash-attn", "--device-map", "cuda:0"], + ["internlm/internlm-xcomposer2-vl-1_8b", "--use-flash-attn", "--device-map", "cuda:0"], ["openbmb/MiniCPM-V-2", "--use-flash-attn", "--device-map", "cuda:0"], ["openbmb/MiniCPM-V", "--use-flash-attn", "--device-map", "cuda:0"], ["llava-hf/bakLlava-v1-hf", "--use-flash-attn", "--device-map", "cuda:0"], diff --git a/model_conf_tests.json b/model_conf_tests.json index 8d895d7..1d23495 100644 --- a/model_conf_tests.json +++ b/model_conf_tests.json @@ -19,6 +19,7 @@ ["internlm/internlm-xcomposer2-4khd-7b", "--use-flash-attn", "--device-map", "cuda:0"], ["internlm/internlm-xcomposer2-7b", "--use-flash-attn", "--device-map", "cuda:0"], ["internlm/internlm-xcomposer2-vl-7b", "--use-flash-attn", "--device-map", "cuda:0"], + ["internlm/internlm-xcomposer2-vl-1_8b", "--use-flash-attn", "--device-map", "cuda:0"], ["openbmb/MiniCPM-V-2", "--use-flash-attn", "--device-map", "cuda:0"], ["openbmb/MiniCPM-V", "--use-flash-attn", "--device-map", "cuda:0"], ["llava-hf/bakLlava-v1-hf", "--use-flash-attn", "--device-map", "cuda:0"], diff --git a/vision-alt.sample.env b/vision-alt.sample.env index 0be64c6..f2f16c3 100644 --- a/vision-alt.sample.env +++ b/vision-alt.sample.env @@ -24,6 +24,7 @@ HF_HOME=hf_home #CLI_COMMAND="python vision.py -m internlm/internlm-xcomposer2-4khd-7b --use-flash-attn --device-map cuda:0" # test pass✅, time: 18.8s, mem: 25.7GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m internlm/internlm-xcomposer2-7b --use-flash-attn --device-map cuda:0" # test pass✅, time: 16.0s, mem: 19.1GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m internlm/internlm-xcomposer2-vl-7b --use-flash-attn --device-map cuda:0" # test pass✅, time: 17.3s, mem: 20.3GB, 8/8 tests passed. +#CLI_COMMAND="python vision.py -m internlm/internlm-xcomposer2-vl-1_8b --use-flash-attn --device-map cuda:0" # test pass✅, time: 5.4s, mem: 7.2GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m openbmb/MiniCPM-V-2 --use-flash-attn --device-map cuda:0" # test pass✅, time: 7.0s, mem: 11.6GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m openbmb/MiniCPM-V --use-flash-attn --device-map cuda:0" # test pass✅, time: 6.5s, mem: 7.8GB, 8/8 tests passed. #CLI_COMMAND="python vision.py -m llava-hf/bakLlava-v1-hf --use-flash-attn --device-map cuda:0" # test fail❌, time: 2.0s, mem: 15.9GB, 0/8 tests passed.