Junwen Tan
Junwen Tan
### Is there an existing issue for this? - [X] I have searched the existing issues ### Current Behavior 请问一下怎么样先用100万条数据微调，再用1万条数据微调？就是说，100条数据微调后，有两种checkpoint，一个是官方的，一个是本地的用100万数据训练的checkpoint，那么怎么利用这两个checkpoint做第二次微调呢？代码要怎么更改？ ### Expected Behavior none ### Steps To Reproduce none ###...
### Is there an existing issue for this? - [X] I have searched the existing issues ### Current Behavior 我直接把train指定为下面这样，主要是CUDA_VISIBLE_DEVICES=0,1,2,3这里改了 PRE_SEQ_LEN=128 LR=2e-2 CUDA_VISIBLE_DEVICES=0,1,2,3 python3 main.py \ --do_train \ --train_file junshi/full_train.json...
### Is there an existing issue for this? - [X] I have searched the existing issues ### Current Behavior 1 ### Expected Behavior 1 ### Steps To Reproduce 1 ###...
I found some mistakes in GraphTSR. Only when you know the ground truth can you make inference in GraphTSR.
对于0906.4747v1.2这条数据 推理结果是这样的：shape = (504,3) tensor([[2.4393e-06, 9.9992e-01, 7.9312e-05], [1.2489e-04, 8.2414e-05, 9.9979e-01], [2.4447e-06, 9.9992e-01, 7.6778e-05], ..., [9.9997e-01, 6.7105e-06, 2.5544e-05], [9.9996e-01, 6.1056e-06, 3.1392e-05], [9.9996e-01, 5.7088e-06, 3.7408e-05]]) 真实标签是这样的：shape = (504,3) [(0, 1, 1), (0,...
在./SciTSR/scitsr/data/utils.py 在推理的时候，事先不知道边，是通过以下函数确定边的，为什么k取20这么大，后期有没有办法删除多余的边？  即使我把k调到2，也是很不准的 预测结果是 tensor([2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 2, 2, 2, 2,...
1.为什么用json2Relations函数把structure里面的json文件转成的rel文件和原有的rel文件不一样？ 2.为什么有些节点之间的关系在rel里面没有？哪些边应该不保留？为什么一行的最后一个节点和下一行的起始节点有边？ 0705.0588v1.1  如图所示，3为第一行最后一个节点，4为第二行第一个节点，怎么会在同一行？ 
测试集中没有rel和json文件夹，可是以下代码会检查rel和json文件夹是否存在，不存在则报错，要怎么解决？ def load_dataset(self, dataset_dir, with_cells, trim=None, debug=False, exts=None): dataset, cells = [], [] if exts is None: exts = ['chunk','rel'] if with_cells: exts.append('json') sub_paths = self.get_sub_paths(dataset_dir, exts, trim=trim) for i,...
RuntimeError: CUDA error: CUBLAS_STATUS_NOT_SUPPORTED when calling `cublasGemmStridedBatchedExFix( handle, opa, opb, m, n, k, (void*)(&falpha), a, CUDA_R_16F, lda, stridea, b, CUDA_R_16F, ldb, strideb, (void*)(&fbeta), c, CUDA_R_16F, ldc, stridec, num_batches, CUDA_R_32F, CUBLAS_GEMM_DEFAULT_TENSOR_OP)`
Could you provide GPU pre-training checkpoints? Run codes with CPU is too slow.