Skip to content

GitLab

Commit 3f891b64 authored by Mufei Li's avatar Mufei Li Committed by Minjie Wang
Browse files

[Tutorial] Fix gat tutorial (#420)

parent 724aa0ca
Hide whitespace changes
Inline Side-by-side
Showing with 5 additions and 6 deletions
tutorials/models/1_gnn/9_gat.py
View file @ 3f891b64
...@@ -474,14 +474,13 @@ for epoch in range(30): ...@@ -474,14 +474,13 @@ for epoch in range(30):
# #
# Again, comparing with uniform distribution: # Again, comparing with uniform distribution:
# #
# .. image:: https://s3.us-east-2.amazonaws.com/dgl.ai/tutorial/gat/ppi-attention-uniform-hist.png # .. image:: https://s3.us-east-2.amazonaws.com/dgl.ai/tutorial/gat/ppi-uniform-hist.png
# :width: 250px # :width: 250px
# :align: center # :align: center
# #
# Clearly, **GAT does learn sharp attention weights**! There is a clear pattern # Clearly, **GAT does learn sharp attention weights**! There is a clear pattern
# over the layers as well: **the attention gets sharper with higher # over the layers as well: **the attention gets sharper with higher
# layer**; the final layer is a mean aggregation thus it is not surprising that # layer**.
# the histogram looks similar to the uniform one.
# #
# Unlike the Cora dataset where GAT's gain is lukewarm at best, for PPI there # Unlike the Cora dataset where GAT's gain is lukewarm at best, for PPI there
# is a significant performance gap between GAT and other GNN variants compared # is a significant performance gap between GAT and other GNN variants compared
...@@ -503,6 +502,6 @@ for epoch in range(30): ...@@ -503,6 +502,6 @@ for epoch in range(30):
# * Stay tune for our next tutorial about how to speedup GAT models by parallelizing multiple attention heads and SPMV optimization. # * Stay tune for our next tutorial about how to speedup GAT models by parallelizing multiple attention heads and SPMV optimization.
# #
# .. |image2| image:: https://s3.us-east-2.amazonaws.com/dgl.ai/tutorial/gat/cora-attention-hist.png # .. |image2| image:: https://s3.us-east-2.amazonaws.com/dgl.ai/tutorial/gat/cora-attention-hist.png
# .. |image5| image:: https://s3.us-east-2.amazonaws.com/dgl.ai/tutorial/gat/ppi-attention-hist-layer1.png # .. |image5| image:: https://s3.us-east-2.amazonaws.com/dgl.ai/tutorial/gat/ppi-first-layer-hist.png
# .. |image6| image:: https://s3.us-east-2.amazonaws.com/dgl.ai/tutorial/gat/ppi-attention-hist-layer2.png # .. |image6| image:: https://s3.us-east-2.amazonaws.com/dgl.ai/tutorial/gat/ppi-second-layer-hist.png
# .. |image7| image:: https://s3.us-east-2.amazonaws.com/dgl.ai/tutorial/gat/ppi-attention-hist-final.png # .. |image7| image:: https://s3.us-east-2.amazonaws.com/dgl.ai/tutorial/gat/ppi-final-layer-hist.png
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment