d6160d4194 cast parameter
zoulq
- Joined on Oct 04, 2022
Loading Heatmap…
zoulq commented on pull request OpenI/MSAdapter#423
Fixed conflict in GPU and Ascendtensor方法不要调用adapter function方法
12 hours ago
zoulq commented on pull request OpenI/MSAdapter#423
Fixed conflict in GPU and Ascendfrom_numpy不要放到tensor.py下面
12 hours ago
zoulq commented on pull request OpenI/MSAdapter#423
Fixed conflict in GPU and Ascend这一句可以放到表格最前面
12 hours ago
zoulq commented on pull request OpenI/MSAdapter#419
nn.MultiheadAttention如果上面调用的multi_head_attention_forward是mindspore接口，这里的attn_output是mindspore tensor吧？
12 hours ago
zoulq pushed to master at OpenI/MSAdapter
2d06953cd9 adapt some ops for ascend (#425) do some change optim i0 adapt i0 adapt svd use from_numpy adapt fmax and fmin Co-authored-by: lvhaoyu <lvhaoyu@huawei.com> Reviewed-on: https://openi.pcl.ac.cn/OpenI/MSAdapter/pulls/425
1 day ago
zoulq pushed to master at OpenI/MSAdapter
1 day ago
zoulq commented on pull request OpenI/MSAdapter#425
adapt some ops for ascend公共接口可以在common目录下建一个文件放
1 day ago
zoulq commented on pull request OpenI/MSAdapter#425
adapt some ops for ascend这里为什么不区分平台？
2 days ago
zoulq commented on pull request OpenI/MSAdapter#425
adapt some ops for ascend中间计算直接调用mindspore的from_numpy，减少cast tensor的转换成本
2 days ago
zoulq commented on pull request OpenI/MSAdapter#419
nn.MultiheadAttention可以直接参考https://github.com/pytorch/pytorch/blob/master/test/test_nn.py里的test_multihead_attn_add_zero_attn等用例进行功能完备性测试
3 days ago
zoulq pushed to master at OpenI/MSAdapter
3 days ago