hurun
hurun
https://github.com/bytedance/lightseq/issues/148 使用pip install lightseq
when i use the below command,i want get the node name before clustering ,how i should program to get it. result = mc.run_mcl(matrix) clusters = mc.get_clusters(result)
Here is setting file, picodet_s_qat_dis and picodet_reader. picodet_s_qat_dis.yaml ----------------------------- Global: reader_config: ./configs/picodet_reader.yml input_list: ['image', 'scale_factor'] Evaluation: True model_dir: ./picodet_s_288_coco_lcnet/ model_filename: model.pdmodel params_filename: model.pdiparams Distillation: alpha: 1.0 loss: l2 Quantization: use_pact:...
@yghstill 已按照提示修改scale_factor，修改后依次出现如下问题，尝试处理但没解决根源问题，还请指点 1、 已将Global改为 Global: input_list: ['image'] 程序能正常运行并计算损失（batchsize=1）， 2022-08-17 10:21:19,109-INFO: Total iter: 980, epoch: 0, batch: 980, loss: [9.578232] 2022-08-17 10:21:20,041-INFO: Total iter: 990, epoch: 0, batch: 990, loss: [14.332893]...
i also stuck this problem,and not solve it still now.
i get the same question,hope someone can help
use this to solve problem https://github.com/google/trax 1 uninstall trax 2 !pip install -q -U trax 3 restart your python enviroment
同问，而且加载lora模型的权重lora_B权重为0
meet the same issue.
meet the same issue. anyone can write a doc or any blog to show how to use colossalai to infer