在Chinese-CLIP中用起来! 在Chinese-CLIP finetune中对于图像端应用知识蒸馏并不复杂。只需要在finetune的sh脚本中加入--distillation配置项。然后在配置项--teacher-model-name填入所要使用的Teacher model名称。现在支持的Teacher model包括以下四种。 Teacher model 模型介绍 damo/multi-modal_clip-vit-huge-patch14_...
//github.com/OFA-Sys/Chinese-CLIP/blob/master/README_En.md#code-organization checkpoint_path=${DATAPATH}/pretrained_weights/clip_cn_vit-b-16.pt # Specify the full path to the ckpt to be converted mkdir -p ${DATAPATH}/deploy/ # Create output folders for ONNX models python cn_clip/...
China's southern city of Shenzhen launched its first Europe-bound freight train on Monday, becoming the 29th Chinese city to introduce rail service to Europe. First Islamic bank opens in Morocco after long-time rejection The first Islamic bank has opened in Morocco after a long time of rejectio...
Applying FlashAttention to the finetune process of Chinese-CLIP is very simple, just add --use-flash-attention to the sh script of finetune. We provide the sample script run_scripts/muge_finetune_vit-b-16_rbt-base_flashattn.sh. Training Speed and Memory Usage Comparison Enabling FlashAttenti...
seven-xu/Chinese-CLIPPublic forked fromOFA-Sys/Chinese-CLIP NotificationsYou must be signed in to change notification settings Fork0 Star0 Files master .github assets cn_clip examples run_scripts .gitignore MIT-LICENSE.txt README.md README_En.md ...
Chinese version of CLIP which achieves Chinese cross-modal retrieval and representation generation. - Chinese-CLIP/README.md at master · seven-xu/Chinese-CLIP
Chinese version of CLIP which achieves Chinese cross-modal retrieval and representation generation. - Chinese-CLIP/zeroshot_dataset.md at master · RainZs/Chinese-CLIP
Chinese version of CLIP which achieves Chinese cross-modal retrieval and representation generation. - Merge remote-tracking branch 'origin/master' into jr_dev · OFA-Sys/Chinese-CLIP@aadaa58
在Chinese-CLIP中用起来! 在Chinese-CLIP finetune中对于图像端应用知识蒸馏并不复杂。只需要在finetune的sh脚本中加入--distllation配置项。然后在配置项--teacher-model-name填入所要使用的Teacher model名称。现在支持的Teacher model包括以下四种。 Teacher model 模型介绍 damo/multi-modal_clip-vit-huge-patch14_...
最新的Chinese-CLIP代码,已支持将各规模的Pytorch模型,转换为ONNX或TensorRT格式,从而相比原始Pytorch模型 提升特征计算的推理速度,同时不影响特征提取的下游任务效果。下面我们给出在GPU上,准备ONNX和TensorRT格式的FP16 Chinese-CLIP部署模型的整个流程(同时给出了Chinese-CLIP预训练TensorRT模型的下载方式),并附上模型效...