denise-k opened a new issue #8608:
URL: https://github.com/apache/tvm/issues/8608


   # Context
   
   CI is taking a while (see #8552 and 
[build#1384](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/)), and I 
am tracking short-term fixes to alleviate this.
   
   # Summary of the issue
   
   `test_lstms.py` was recently added in PR #8447. It takes 20+ minutes to run 
8 network-sized tests.
   
   The tests uses the following parameters, which don't seem minimal: 
   
   ```
   ## Model parameters
   model_feature_size = 5
   model_hidden_size = 10
   model_num_layers = 2
   seqs_length = 15
   projection_size = 7
   batch_size = 3
   ```
   
   Also, the In the test console, there are 64 of the following warnings. It 
seems like there are issues exporting the pytorch frontend to ONNX:
   ```
   WARNING: projection is not supported for torch version less than 1.8.0!  
LSTM was constructed without projection!
   WARNING: torch.onnx.export does not support conversion LSTM with projection 
from pytorch! TODO: waiting for the support and correct test after that.
   ```
   
   # Suggested actions
   
   * Try to minimize the size of the model testcases even further.
   * Consider minimizing the number of calls to `torch.onnx.export` until we 
get a nightly CI in place.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to