-
Notifications
You must be signed in to change notification settings - Fork 42
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
baseline_vs.embeddings #21
Comments
You can get the embeddings here: http://rtw.ml.cmu.edu/emnlp2015_sfe/embeddings.tgz. Extract that and change the relevant paths in the json file, and let me know if you run into any more issues. |
After solving a bunch of problems of unexpected key such as "mode", "path selector" and the like, I run into a problem concerning scala grammar. In file "FeatureGenerator.scala"(line 85: new PraFeatureGenerator), IDEA said that "Expression of type PraFeatureGenerator doesn't conform to expected type FeatureGenerator[T]". And this problem causes failure to compile. |
Did you change any code? You should not have to change any code, only the configuration files. |
I added the embeddings file. And the error occured |
The problem here is that I changed the code to crash if you request an edge id where the edge isn't in the graph, and I haven't run this particular part of the code since I made that change, so I never discovered this bug. The old behavior was to return -1, which worked fine here, but was undesirable for other reasons. To fix this, you need to add a check here for whether the edge is in the graph (I believe the correct method is |
After fixing this, the code seems to run. Like running the SFE which keeps outputting "[error] getValue", running pra keeps outputting "[error] engine execUpdates - t:1 INFO: Waiting for execution to finish: countDown:1". My machine only has 8GB RAM. And I set "javaOptions in run ++= Seq("-Xmx4g")". As a result, java heap space outOfMemoryError. What is the least space to satisfy the program? |
Sadly, the code needs a whole lot of memory to run; see the note on the main README. |
Thanks, Matt. I set "Xmx5g". Although it ends up with out of memory again, some relations still succeed to generate "scores.tsv" and "weights.tsv". So, I think maybe I can run only one relation at a time. Eventually, I can have all the 10 relations run. I will pull a request as soon as the programming runs steady. |
If I want to run original pra algo with different dataset , can I use the same embedding file as you posted before? |
You almost certainly want different embeddings in that case. I just got them from running PCA. There's a way to do that with this code, if you want to dig through it and figure it out, or you could just use Matlab or something instead. |
Hi, Matt. I was trying to run PRA instead of SFE. But I run into the following problem.
[error] Exception in thread "main" java.lang.IllegalStateException: Error specifying embeddings (you must give full paths to embedding files)
So, I want to know what does "pca_svo" mean in file baseline_vs*.json.
The text was updated successfully, but these errors were encountered: