search

juexinwang / scGNN

scGNN (single cell graph neural networks) for single cell clustering and imputation using graph neural networks
MIT License
134 stars 40 forks source link

执行scGNN.py时候，关于cublasSgemm矩阵乘法，我出现了RuntimeError，该怎么解决呢？就是执行最后一段代码 #32

Open joyyy111 opened 8 months ago

joyyy111 commented 8 months ago

1711000756751

juexinwang commented 8 months ago

You may need to use the specific version of python/numpy/... as recommended

On Thu, Mar 21, 2024 at 2:00 AM joyyy111 @.***> wrote:

1711000756751.jpg (view on web) https://github.com/juexinwang/scGNN/assets/138951832/aefbf37f-c9a9-4539-a98c-d95a8ac93526

— Reply to this email directly, view it on GitHub https://github.com/juexinwang/scGNN/issues/32, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZPXTECII7HAYODSHNEOR3YZJZOXAVCNFSM6AAAAABFAZG6WKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGE4TSMRQGI4DMMA . You are receiving this because you are subscribed to this thread.Message ID: @.***>

wwhgh commented 7 months ago

I also encountered this issue, while the verson of numpy is 1.18.1.

I followed the scripts in the REPRODUCIBILITY.md and directly used the preproceeed data. The debug info as follows:

---0:00:00---scRNA starts loading. ---0:00:01---scRNA has been successfully loaded Mem consumption: 1610100 ---0:00:01---Start feature autoencoder training Mem consumption: 1610356 Traceback (most recent call last): File "main_benchmark.py", line 384, in recon, original, z = train(epoch, EMFlag=False) File "main_benchmark.py", line 291, in train recon_batch, z = model(data) ... ... RuntimeError: CUDA error: CUBLAS_STATUS_EXECUTION_FAILED when calling cublasSgemm( handle, opa, opb, m, n, k, &alpha, a, lda, b, ldb, &beta, c, ldc)

How to fixed that runtime error ? Thanks a lot

juexinwang commented 7 months ago

Hi, did you use all the recommended versions of numpy and python?

On Wed, Mar 27, 2024 at 7:38 AM Weihua Wang @.***> wrote:

I also encountered this issue, while the verson of numpy is 1.18.1.

I followed the scripts in the REPRODUCIBILITY.md and directly used the preproceeed data. The debug info as follows:

---0:00:00---scRNA starts loading. ---0:00:01---scRNA has been successfully loaded Mem consumption: 1610100 ---0:00:01---Start feature autoencoder training Mem consumption: 1610356 Traceback (most recent call last): File "main_benchmark.py", line 384, in recon, original, z = train(epoch, EMFlag=False) File "main_benchmark.py", line 291, in train recon_batch, z = model(data) ... ... RuntimeError: CUDA error: CUBLAS_STATUS_EXECUTION_FAILED when calling cublasSgemm( handle, opa, opb, m, n, k, &alpha, a, lda, b, ldb, &beta, c, ldc)

How to fixed that runtime error ? Thanks a lot

— Reply to this email directly, view it on GitHub https://github.com/juexinwang/scGNN/issues/32#issuecomment-2022560420, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAZPXTAVIJALBFJMXWHNIJDY2KVTHAVCNFSM6AAAAABFAZG6WKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAMRSGU3DANBSGA . You are receiving this because you commented.Message ID: @.***>

moomight commented 5 months ago

I happened to encounter this situation as well, have you solved this problem? Thanks for your response.

moomight commented 5 months ago

I happened to encounter this situation as well, have you solved this problem? Thanks for your response.

Well, I have solved this issue. Just set torch to use CPU instead of GPU. I write export CUDA_VISIBLE_DEVICES="" in terminal, and it works.