You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2020/01/08 05:17:06 UTC
[GitHub] [incubator-tvm] yongfeng-nv opened a new pull request #4651: Tensor
Expression Debug Display (TEDD)
yongfeng-nv opened a new pull request #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651
This is the PR for a Tensor Expression visualization tool ([RFC](https://discuss.tvm.ai/t/visualize-tensor-expression/5174)). Any feedback or comment are welcome.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on issue #4651: Tensor
Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-587540016
@tqchen, this PR has been clear for a week. It still needs uploading three static images for the tutorial. Can you help moving it forward? Thank you.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] tqchen edited a comment on issue #4651: Tensor
Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
tqchen edited a comment on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-573377458
Thanks @yongfeng-nv . One thing that I think worth considering, as in many viz tools, is the separation of visualization data schema(in this case the perhaps a dom tree or similar kind) from the visualization itself(graphviz).
We can have a tool that extracts the spec into json, then have a tool to take that spec and visualize it
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on issue #4651: Tensor
Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-571895493
@Hzfengsy, could you help reviewing this PR?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on a change in pull request
#4651: Tensor Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on a change in pull request #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#discussion_r365005585
##########
File path: python/tvm/contrib/tedd.py
##########
@@ -0,0 +1,506 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Tensor Expression Debug Display (TEDD), visualizing Tensor Expression"""
+import html
+from graphviz import Digraph
+from graphviz import Source
+from IPython.display import display
+from IPython.display import SVG
+import tvm
+
+TVMDD_TABLE_BODY_WIDTH = 30
+# Must match enum IterVarType defined in include/tvm/expr.h
+ITERVAR_TYPE_STRING_MAP = {
+ 0: ('kDataPar', '#FFFFFF'),
+ 1: ('kThreadIndex', '#2980B9'),
+ 2: ('kCommReduce', '#FAD7A0'),
+ 3: ('kOrdered', '#D35400'),
+ 4: ('kOpaque', '#ABB2B9'),
+ 5: ('kUnrolled', '#D2B4DE'),
+ 6: ('kVectorized', '#AED6F1'),
+ 7: ('kParallelized', '#F5B7B1'),
+ 8: ('kTensorized', '#A9DFBF'),
+}
+
+
+def get_or_create_dot_id(obj, prefix="", assert_on_missing=False):
+ """If obj's ID has been registered, return it.
+ If not, either assert or create a unique and legal ID, register and
+ return it, according to assert_on_missing.
+ ID must be a unique and legal Dotty ID.
+
+ Parameters
+ ----------
+ obj : objet
+ Serve as the key to the ID.
+
+ prefix : string
+ Prefix to attach to the ID. Usually use obj's non-unique
+ name as prefix.
+
+ assert_on_missing : bool
+ Asserts or not if object doesn't have a registered ID.
+ """
+ prefix = prefix.replace('.', '_')
+ if not hasattr(get_or_create_dot_id, "obj_id_dict"):
+ get_or_create_dot_id.obj_id_dict = {}
+ if obj not in get_or_create_dot_id.obj_id_dict:
+ if assert_on_missing:
+ assert False, 'dot_id ' + str(obj) + ' has not been registered.'
+ else:
+ get_or_create_dot_id.obj_id_dict[obj] = prefix + hex(id(obj))
+ return get_or_create_dot_id.obj_id_dict[obj]
+
+
+def get_port_id(is_input, index):
+ return 'I_' + str(index) if is_input else 'O_' + str(index)
+
+
+def get_itervar_type_info(iter_type):
+ assert iter_type < len(
+ ITERVAR_TYPE_STRING_MAP), 'Unknown IterVar type: ' + str(iter_type)
+ return ITERVAR_TYPE_STRING_MAP[iter_type]
+
+
+def get_itervar_label_color(itervar, iv_type):
+ type_info = get_itervar_type_info(iv_type)
+ return str(itervar.var) + '(' + type_info[0] + ')', type_info[1]
+
+
+def linebrk(s, n):
+ """ Break input string s with <br/> for every n charactors."""
+ result = ''
+ j = 0
+ for i, c in enumerate(s):
+ if j == n and i != len(s) - 1:
+ result = result + '\n'
+ j = 0
+ j = j + 1
+ result = result + c
+ result = html.escape(str(result), quote=True)
+ result = result.replace('\n', '<br/>')
+ return result
+
+
+def create_graph(name="", rankdir='BT'):
+ graph = Digraph(name=name)
+ graph.graph_attr['rankdir'] = rankdir
+ return graph
+
+
+def itervar_label(itervar, index, index_color, label):
+ return '<TR><TD PORT="' + get_or_create_dot_id(itervar, str(
+ itervar.var)) + '" BGCOLOR="' + index_color + '">' + str(
+ index
+ ) + '</TD><TD BGCOLOR="white" PORT="itervar">' + label + '</TD></TR>'
+
+
+def stage_label(stage):
+ return stage.op.name + '<br/>Scope: ' + stage.scope
+
+
+def legend_label():
+ label = '<<TABLE BORDER="0" CELLBORDER="1" CELLSPACING="0" CELLPADDING="4">'
+ for iter_type in ITERVAR_TYPE_STRING_MAP:
+ name, color = ITERVAR_TYPE_STRING_MAP[iter_type]
+ label += '<TR><TD BGCOLOR="' + color + '"></TD>' \
+ + '<TD BGCOLOR="white">' + name + '</TD></TR>'
+ label += '</TABLE>>'
+ return label
+
+
+def legend_dot(g):
+ with g.subgraph(name='cluster_legend') as subgraph:
+ subgraph.attr(label='Legend')
+ label = legend_label()
+ subgraph.node('legend', label, shape='none', margin='0')
+
+
+def dump_graph(dot_string,
+ showsvg=True,
+ dotfilepath='',
+ outputdotstring=False):
+ """Output dot_string in various formats."""
+ if dotfilepath:
+ try:
+ dot_file = open(dotfilepath, "w+")
+ dot_file.write(dot_string)
+ dot_file.close()
+ except IOError:
+ print('Cannot open file: ' + dotfilepath)
+ if showsvg:
+ src = Source(dot_string)
+ display(SVG(src.pipe(format='svg')))
+ if outputdotstring:
+ return dot_string
+ return None
+
+
+def viz_schedule_tree(sch,
+ showsvg=False,
Review comment:
@Hzfengsy, I have updated the variable names and avoided some lint warnings from the tests. But the test will still fail because of the ci-cpu doesn't have graphviz. I haven't seen response to my question below. Do you have any suggestion? Thanks
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] tqchen commented on issue #4651: Tensor Expression
Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
tqchen commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-573377458
Thanks @yongfeng-nv . One thing that I think worth considering, as in many viz tools, is the separation of visualization data source specification(in this case the perhaps a dom tree or similar kind) from the visualization(graphviz).
We can have a tool that extracts the spec into json, then have a tool to take that spec and visualize it
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on issue #4651: Tensor
Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-574416251
> Thanks @yongfeng-nv . One thing that I think worth considering, as in many viz tools, is the separation of visualization data source specification(in this case the perhaps a dom tree or similar kind) from the visualization(graphviz).
>
> We can have a tool that extracts the spec into json, then have a tool to take that spec and visualize it
@tqchen, I see your point. Let me spec the data source spec.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] Hzfengsy commented on issue #4651: Tensor
Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
Hzfengsy commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-584170658
Thank you for the update.
I think the location on `dmlc/web-data` is good, but maybe need others' help to commit.
You can render the tutorial at local by
```
python3 -m pip install sphinx-gallery
cd tvm_path/docs
make html
```
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on a change in pull request
#4651: Tensor Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on a change in pull request #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#discussion_r365031269
##########
File path: python/tvm/contrib/tedd.py
##########
@@ -0,0 +1,506 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Tensor Expression Debug Display (TEDD), visualizing Tensor Expression"""
+import html
+from graphviz import Digraph
+from graphviz import Source
+from IPython.display import display
+from IPython.display import SVG
+import tvm
+
+TVMDD_TABLE_BODY_WIDTH = 30
+# Must match enum IterVarType defined in include/tvm/expr.h
+ITERVAR_TYPE_STRING_MAP = {
+ 0: ('kDataPar', '#FFFFFF'),
+ 1: ('kThreadIndex', '#2980B9'),
+ 2: ('kCommReduce', '#FAD7A0'),
+ 3: ('kOrdered', '#D35400'),
+ 4: ('kOpaque', '#ABB2B9'),
+ 5: ('kUnrolled', '#D2B4DE'),
+ 6: ('kVectorized', '#AED6F1'),
+ 7: ('kParallelized', '#F5B7B1'),
+ 8: ('kTensorized', '#A9DFBF'),
+}
+
+
+def get_or_create_dot_id(obj, prefix="", assert_on_missing=False):
+ """If obj's ID has been registered, return it.
+ If not, either assert or create a unique and legal ID, register and
+ return it, according to assert_on_missing.
+ ID must be a unique and legal Dotty ID.
+
+ Parameters
+ ----------
+ obj : objet
+ Serve as the key to the ID.
+
+ prefix : string
+ Prefix to attach to the ID. Usually use obj's non-unique
+ name as prefix.
+
+ assert_on_missing : bool
+ Asserts or not if object doesn't have a registered ID.
+ """
+ prefix = prefix.replace('.', '_')
+ if not hasattr(get_or_create_dot_id, "obj_id_dict"):
+ get_or_create_dot_id.obj_id_dict = {}
+ if obj not in get_or_create_dot_id.obj_id_dict:
+ if assert_on_missing:
+ assert False, 'dot_id ' + str(obj) + ' has not been registered.'
+ else:
+ get_or_create_dot_id.obj_id_dict[obj] = prefix + hex(id(obj))
+ return get_or_create_dot_id.obj_id_dict[obj]
+
+
+def get_port_id(is_input, index):
+ return 'I_' + str(index) if is_input else 'O_' + str(index)
+
+
+def get_itervar_type_info(iter_type):
+ assert iter_type < len(
+ ITERVAR_TYPE_STRING_MAP), 'Unknown IterVar type: ' + str(iter_type)
+ return ITERVAR_TYPE_STRING_MAP[iter_type]
+
+
+def get_itervar_label_color(itervar, iv_type):
+ type_info = get_itervar_type_info(iv_type)
+ return str(itervar.var) + '(' + type_info[0] + ')', type_info[1]
+
+
+def linebrk(s, n):
+ """ Break input string s with <br/> for every n charactors."""
+ result = ''
+ j = 0
+ for i, c in enumerate(s):
+ if j == n and i != len(s) - 1:
+ result = result + '\n'
+ j = 0
+ j = j + 1
+ result = result + c
+ result = html.escape(str(result), quote=True)
+ result = result.replace('\n', '<br/>')
+ return result
+
+
+def create_graph(name="", rankdir='BT'):
+ graph = Digraph(name=name)
+ graph.graph_attr['rankdir'] = rankdir
+ return graph
+
+
+def itervar_label(itervar, index, index_color, label):
+ return '<TR><TD PORT="' + get_or_create_dot_id(itervar, str(
+ itervar.var)) + '" BGCOLOR="' + index_color + '">' + str(
+ index
+ ) + '</TD><TD BGCOLOR="white" PORT="itervar">' + label + '</TD></TR>'
+
+
+def stage_label(stage):
+ return stage.op.name + '<br/>Scope: ' + stage.scope
+
+
+def legend_label():
+ label = '<<TABLE BORDER="0" CELLBORDER="1" CELLSPACING="0" CELLPADDING="4">'
+ for iter_type in ITERVAR_TYPE_STRING_MAP:
+ name, color = ITERVAR_TYPE_STRING_MAP[iter_type]
+ label += '<TR><TD BGCOLOR="' + color + '"></TD>' \
+ + '<TD BGCOLOR="white">' + name + '</TD></TR>'
+ label += '</TABLE>>'
+ return label
+
+
+def legend_dot(g):
+ with g.subgraph(name='cluster_legend') as subgraph:
+ subgraph.attr(label='Legend')
+ label = legend_label()
+ subgraph.node('legend', label, shape='none', margin='0')
+
+
+def dump_graph(dot_string,
+ showsvg=True,
+ dotfilepath='',
+ outputdotstring=False):
+ """Output dot_string in various formats."""
+ if dotfilepath:
+ try:
+ dot_file = open(dotfilepath, "w+")
+ dot_file.write(dot_string)
+ dot_file.close()
+ except IOError:
+ print('Cannot open file: ' + dotfilepath)
+ if showsvg:
+ src = Source(dot_string)
+ display(SVG(src.pipe(format='svg')))
+ if outputdotstring:
+ return dot_string
+ return None
+
+
+def viz_schedule_tree(sch,
+ showsvg=False,
Review comment:
I am also working on a tutorial. How about putting it under tutorial/language?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] tqchen commented on issue #4651: Tensor Expression
Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
tqchen commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-591231104
Thanks @yongfeng-nv @Hzfengsy this PR is now merged!
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] Hzfengsy commented on a change in pull request
#4651: Tensor Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
Hzfengsy commented on a change in pull request #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#discussion_r364087171
##########
File path: python/tvm/contrib/tedd.py
##########
@@ -0,0 +1,506 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Tensor Expression Debug Display (TEDD), visualizing Tensor Expression"""
+import html
+from graphviz import Digraph
+from graphviz import Source
+from IPython.display import display
+from IPython.display import SVG
+import tvm
+
+TVMDD_TABLE_BODY_WIDTH = 30
+# Must match enum IterVarType defined in include/tvm/expr.h
+ITERVAR_TYPE_STRING_MAP = {
+ 0: ('kDataPar', '#FFFFFF'),
+ 1: ('kThreadIndex', '#2980B9'),
+ 2: ('kCommReduce', '#FAD7A0'),
+ 3: ('kOrdered', '#D35400'),
+ 4: ('kOpaque', '#ABB2B9'),
+ 5: ('kUnrolled', '#D2B4DE'),
+ 6: ('kVectorized', '#AED6F1'),
+ 7: ('kParallelized', '#F5B7B1'),
+ 8: ('kTensorized', '#A9DFBF'),
+}
+
+
+def get_or_create_dot_id(obj, prefix="", assert_on_missing=False):
+ """If obj's ID has been registered, return it.
+ If not, either assert or create a unique and legal ID, register and
+ return it, according to assert_on_missing.
+ ID must be a unique and legal Dotty ID.
+
+ Parameters
+ ----------
+ obj : objet
+ Serve as the key to the ID.
+
+ prefix : string
+ Prefix to attach to the ID. Usually use obj's non-unique
+ name as prefix.
+
+ assert_on_missing : bool
+ Asserts or not if object doesn't have a registered ID.
+ """
+ prefix = prefix.replace('.', '_')
+ if not hasattr(get_or_create_dot_id, "obj_id_dict"):
+ get_or_create_dot_id.obj_id_dict = {}
+ if obj not in get_or_create_dot_id.obj_id_dict:
+ if assert_on_missing:
+ assert False, 'dot_id ' + str(obj) + ' has not been registered.'
+ else:
+ get_or_create_dot_id.obj_id_dict[obj] = prefix + hex(id(obj))
+ return get_or_create_dot_id.obj_id_dict[obj]
+
+
+def get_port_id(is_input, index):
+ return 'I_' + str(index) if is_input else 'O_' + str(index)
+
+
+def get_itervar_type_info(iter_type):
+ assert iter_type < len(
+ ITERVAR_TYPE_STRING_MAP), 'Unknown IterVar type: ' + str(iter_type)
+ return ITERVAR_TYPE_STRING_MAP[iter_type]
+
+
+def get_itervar_label_color(itervar, iv_type):
+ type_info = get_itervar_type_info(iv_type)
+ return str(itervar.var) + '(' + type_info[0] + ')', type_info[1]
+
+
+def linebrk(s, n):
+ """ Break input string s with <br/> for every n charactors."""
+ result = ''
+ j = 0
+ for i, c in enumerate(s):
+ if j == n and i != len(s) - 1:
+ result = result + '\n'
+ j = 0
+ j = j + 1
+ result = result + c
+ result = html.escape(str(result), quote=True)
+ result = result.replace('\n', '<br/>')
+ return result
+
+
+def create_graph(name="", rankdir='BT'):
+ graph = Digraph(name=name)
+ graph.graph_attr['rankdir'] = rankdir
+ return graph
+
+
+def itervar_label(itervar, index, index_color, label):
+ return '<TR><TD PORT="' + get_or_create_dot_id(itervar, str(
+ itervar.var)) + '" BGCOLOR="' + index_color + '">' + str(
+ index
+ ) + '</TD><TD BGCOLOR="white" PORT="itervar">' + label + '</TD></TR>'
+
+
+def stage_label(stage):
+ return stage.op.name + '<br/>Scope: ' + stage.scope
+
+
+def legend_label():
+ label = '<<TABLE BORDER="0" CELLBORDER="1" CELLSPACING="0" CELLPADDING="4">'
+ for iter_type in ITERVAR_TYPE_STRING_MAP:
+ name, color = ITERVAR_TYPE_STRING_MAP[iter_type]
+ label += '<TR><TD BGCOLOR="' + color + '"></TD>' \
+ + '<TD BGCOLOR="white">' + name + '</TD></TR>'
+ label += '</TABLE>>'
+ return label
+
+
+def legend_dot(g):
+ with g.subgraph(name='cluster_legend') as subgraph:
+ subgraph.attr(label='Legend')
+ label = legend_label()
+ subgraph.node('legend', label, shape='none', margin='0')
+
+
+def dump_graph(dot_string,
+ showsvg=True,
+ dotfilepath='',
+ outputdotstring=False):
+ """Output dot_string in various formats."""
+ if dotfilepath:
+ try:
+ dot_file = open(dotfilepath, "w+")
+ dot_file.write(dot_string)
+ dot_file.close()
+ except IOError:
+ print('Cannot open file: ' + dotfilepath)
+ if showsvg:
+ src = Source(dot_string)
+ display(SVG(src.pipe(format='svg')))
+ if outputdotstring:
+ return dot_string
+ return None
+
+
+def viz_schedule_tree(sch,
+ showsvg=False,
Review comment:
Can you please change the argument name to `show_svg`, `dot_file_path` and `output_dot_string`?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] Hzfengsy commented on a change in pull request
#4651: Tensor Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
Hzfengsy commented on a change in pull request #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#discussion_r365149671
##########
File path: python/tvm/contrib/tedd.py
##########
@@ -0,0 +1,506 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Tensor Expression Debug Display (TEDD), visualizing Tensor Expression"""
+import html
+from graphviz import Digraph
+from graphviz import Source
+from IPython.display import display
+from IPython.display import SVG
+import tvm
+
+TVMDD_TABLE_BODY_WIDTH = 30
+# Must match enum IterVarType defined in include/tvm/expr.h
+ITERVAR_TYPE_STRING_MAP = {
+ 0: ('kDataPar', '#FFFFFF'),
+ 1: ('kThreadIndex', '#2980B9'),
+ 2: ('kCommReduce', '#FAD7A0'),
+ 3: ('kOrdered', '#D35400'),
+ 4: ('kOpaque', '#ABB2B9'),
+ 5: ('kUnrolled', '#D2B4DE'),
+ 6: ('kVectorized', '#AED6F1'),
+ 7: ('kParallelized', '#F5B7B1'),
+ 8: ('kTensorized', '#A9DFBF'),
+}
+
+
+def get_or_create_dot_id(obj, prefix="", assert_on_missing=False):
+ """If obj's ID has been registered, return it.
+ If not, either assert or create a unique and legal ID, register and
+ return it, according to assert_on_missing.
+ ID must be a unique and legal Dotty ID.
+
+ Parameters
+ ----------
+ obj : objet
+ Serve as the key to the ID.
+
+ prefix : string
+ Prefix to attach to the ID. Usually use obj's non-unique
+ name as prefix.
+
+ assert_on_missing : bool
+ Asserts or not if object doesn't have a registered ID.
+ """
+ prefix = prefix.replace('.', '_')
+ if not hasattr(get_or_create_dot_id, "obj_id_dict"):
+ get_or_create_dot_id.obj_id_dict = {}
+ if obj not in get_or_create_dot_id.obj_id_dict:
+ if assert_on_missing:
+ assert False, 'dot_id ' + str(obj) + ' has not been registered.'
+ else:
+ get_or_create_dot_id.obj_id_dict[obj] = prefix + hex(id(obj))
+ return get_or_create_dot_id.obj_id_dict[obj]
+
+
+def get_port_id(is_input, index):
+ return 'I_' + str(index) if is_input else 'O_' + str(index)
+
+
+def get_itervar_type_info(iter_type):
+ assert iter_type < len(
+ ITERVAR_TYPE_STRING_MAP), 'Unknown IterVar type: ' + str(iter_type)
+ return ITERVAR_TYPE_STRING_MAP[iter_type]
+
+
+def get_itervar_label_color(itervar, iv_type):
+ type_info = get_itervar_type_info(iv_type)
+ return str(itervar.var) + '(' + type_info[0] + ')', type_info[1]
+
+
+def linebrk(s, n):
+ """ Break input string s with <br/> for every n charactors."""
+ result = ''
+ j = 0
+ for i, c in enumerate(s):
+ if j == n and i != len(s) - 1:
+ result = result + '\n'
+ j = 0
+ j = j + 1
+ result = result + c
+ result = html.escape(str(result), quote=True)
+ result = result.replace('\n', '<br/>')
+ return result
+
+
+def create_graph(name="", rankdir='BT'):
+ graph = Digraph(name=name)
+ graph.graph_attr['rankdir'] = rankdir
+ return graph
+
+
+def itervar_label(itervar, index, index_color, label):
+ return '<TR><TD PORT="' + get_or_create_dot_id(itervar, str(
+ itervar.var)) + '" BGCOLOR="' + index_color + '">' + str(
+ index
+ ) + '</TD><TD BGCOLOR="white" PORT="itervar">' + label + '</TD></TR>'
+
+
+def stage_label(stage):
+ return stage.op.name + '<br/>Scope: ' + stage.scope
+
+
+def legend_label():
+ label = '<<TABLE BORDER="0" CELLBORDER="1" CELLSPACING="0" CELLPADDING="4">'
+ for iter_type in ITERVAR_TYPE_STRING_MAP:
+ name, color = ITERVAR_TYPE_STRING_MAP[iter_type]
+ label += '<TR><TD BGCOLOR="' + color + '"></TD>' \
+ + '<TD BGCOLOR="white">' + name + '</TD></TR>'
+ label += '</TABLE>>'
+ return label
+
+
+def legend_dot(g):
+ with g.subgraph(name='cluster_legend') as subgraph:
+ subgraph.attr(label='Legend')
+ label = legend_label()
+ subgraph.node('legend', label, shape='none', margin='0')
+
+
+def dump_graph(dot_string,
+ showsvg=True,
+ dotfilepath='',
+ outputdotstring=False):
+ """Output dot_string in various formats."""
+ if dotfilepath:
+ try:
+ dot_file = open(dotfilepath, "w+")
+ dot_file.write(dot_string)
+ dot_file.close()
+ except IOError:
+ print('Cannot open file: ' + dotfilepath)
+ if showsvg:
+ src = Source(dot_string)
+ display(SVG(src.pipe(format='svg')))
+ if outputdotstring:
+ return dot_string
+ return None
+
+
+def viz_schedule_tree(sch,
+ showsvg=False,
Review comment:
Sounds great! As for the CI test, I would like to skip this test on the CI until the CI updating.
Also, is there a way to allow contributors to modify the docker image on CI? @tqchen
Many new features depend on different environments. It is difficult to change the CI docker every time by only one person.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on a change in pull request
#4651: Tensor Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on a change in pull request #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#discussion_r364318828
##########
File path: python/tvm/contrib/tedd.py
##########
@@ -0,0 +1,506 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Tensor Expression Debug Display (TEDD), visualizing Tensor Expression"""
+import html
+from graphviz import Digraph
+from graphviz import Source
+from IPython.display import display
+from IPython.display import SVG
+import tvm
+
+TVMDD_TABLE_BODY_WIDTH = 30
+# Must match enum IterVarType defined in include/tvm/expr.h
+ITERVAR_TYPE_STRING_MAP = {
+ 0: ('kDataPar', '#FFFFFF'),
+ 1: ('kThreadIndex', '#2980B9'),
+ 2: ('kCommReduce', '#FAD7A0'),
+ 3: ('kOrdered', '#D35400'),
+ 4: ('kOpaque', '#ABB2B9'),
+ 5: ('kUnrolled', '#D2B4DE'),
+ 6: ('kVectorized', '#AED6F1'),
+ 7: ('kParallelized', '#F5B7B1'),
+ 8: ('kTensorized', '#A9DFBF'),
+}
+
+
+def get_or_create_dot_id(obj, prefix="", assert_on_missing=False):
+ """If obj's ID has been registered, return it.
+ If not, either assert or create a unique and legal ID, register and
+ return it, according to assert_on_missing.
+ ID must be a unique and legal Dotty ID.
+
+ Parameters
+ ----------
+ obj : objet
+ Serve as the key to the ID.
+
+ prefix : string
+ Prefix to attach to the ID. Usually use obj's non-unique
+ name as prefix.
+
+ assert_on_missing : bool
+ Asserts or not if object doesn't have a registered ID.
+ """
+ prefix = prefix.replace('.', '_')
+ if not hasattr(get_or_create_dot_id, "obj_id_dict"):
+ get_or_create_dot_id.obj_id_dict = {}
+ if obj not in get_or_create_dot_id.obj_id_dict:
+ if assert_on_missing:
+ assert False, 'dot_id ' + str(obj) + ' has not been registered.'
+ else:
+ get_or_create_dot_id.obj_id_dict[obj] = prefix + hex(id(obj))
+ return get_or_create_dot_id.obj_id_dict[obj]
+
+
+def get_port_id(is_input, index):
+ return 'I_' + str(index) if is_input else 'O_' + str(index)
+
+
+def get_itervar_type_info(iter_type):
+ assert iter_type < len(
+ ITERVAR_TYPE_STRING_MAP), 'Unknown IterVar type: ' + str(iter_type)
+ return ITERVAR_TYPE_STRING_MAP[iter_type]
+
+
+def get_itervar_label_color(itervar, iv_type):
+ type_info = get_itervar_type_info(iv_type)
+ return str(itervar.var) + '(' + type_info[0] + ')', type_info[1]
+
+
+def linebrk(s, n):
+ """ Break input string s with <br/> for every n charactors."""
+ result = ''
+ j = 0
+ for i, c in enumerate(s):
+ if j == n and i != len(s) - 1:
+ result = result + '\n'
+ j = 0
+ j = j + 1
+ result = result + c
+ result = html.escape(str(result), quote=True)
+ result = result.replace('\n', '<br/>')
+ return result
+
+
+def create_graph(name="", rankdir='BT'):
+ graph = Digraph(name=name)
+ graph.graph_attr['rankdir'] = rankdir
+ return graph
+
+
+def itervar_label(itervar, index, index_color, label):
+ return '<TR><TD PORT="' + get_or_create_dot_id(itervar, str(
+ itervar.var)) + '" BGCOLOR="' + index_color + '">' + str(
+ index
+ ) + '</TD><TD BGCOLOR="white" PORT="itervar">' + label + '</TD></TR>'
+
+
+def stage_label(stage):
+ return stage.op.name + '<br/>Scope: ' + stage.scope
+
+
+def legend_label():
+ label = '<<TABLE BORDER="0" CELLBORDER="1" CELLSPACING="0" CELLPADDING="4">'
+ for iter_type in ITERVAR_TYPE_STRING_MAP:
+ name, color = ITERVAR_TYPE_STRING_MAP[iter_type]
+ label += '<TR><TD BGCOLOR="' + color + '"></TD>' \
+ + '<TD BGCOLOR="white">' + name + '</TD></TR>'
+ label += '</TABLE>>'
+ return label
+
+
+def legend_dot(g):
+ with g.subgraph(name='cluster_legend') as subgraph:
+ subgraph.attr(label='Legend')
+ label = legend_label()
+ subgraph.node('legend', label, shape='none', margin='0')
+
+
+def dump_graph(dot_string,
+ showsvg=True,
+ dotfilepath='',
+ outputdotstring=False):
+ """Output dot_string in various formats."""
+ if dotfilepath:
+ try:
+ dot_file = open(dotfilepath, "w+")
+ dot_file.write(dot_string)
+ dot_file.close()
+ except IOError:
+ print('Cannot open file: ' + dotfilepath)
+ if showsvg:
+ src = Source(dot_string)
+ display(SVG(src.pipe(format='svg')))
+ if outputdotstring:
+ return dot_string
+ return None
+
+
+def viz_schedule_tree(sch,
+ showsvg=False,
Review comment:
I am making these changes.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] tqchen merged pull request #4651: Tensor Expression
Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
tqchen merged pull request #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] tqchen commented on issue #4651: Tensor Expression
Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
tqchen commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-573377518
re CI It would be great to add a testing function to see if graphviz exists, if not skip the test
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on a change in pull request
#4651: Tensor Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on a change in pull request #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#discussion_r365324499
##########
File path: python/tvm/contrib/tedd.py
##########
@@ -0,0 +1,506 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+"""Tensor Expression Debug Display (TEDD), visualizing Tensor Expression"""
+import html
+from graphviz import Digraph
+from graphviz import Source
+from IPython.display import display
+from IPython.display import SVG
+import tvm
+
+TVMDD_TABLE_BODY_WIDTH = 30
+# Must match enum IterVarType defined in include/tvm/expr.h
+ITERVAR_TYPE_STRING_MAP = {
+ 0: ('kDataPar', '#FFFFFF'),
+ 1: ('kThreadIndex', '#2980B9'),
+ 2: ('kCommReduce', '#FAD7A0'),
+ 3: ('kOrdered', '#D35400'),
+ 4: ('kOpaque', '#ABB2B9'),
+ 5: ('kUnrolled', '#D2B4DE'),
+ 6: ('kVectorized', '#AED6F1'),
+ 7: ('kParallelized', '#F5B7B1'),
+ 8: ('kTensorized', '#A9DFBF'),
+}
+
+
+def get_or_create_dot_id(obj, prefix="", assert_on_missing=False):
+ """If obj's ID has been registered, return it.
+ If not, either assert or create a unique and legal ID, register and
+ return it, according to assert_on_missing.
+ ID must be a unique and legal Dotty ID.
+
+ Parameters
+ ----------
+ obj : objet
+ Serve as the key to the ID.
+
+ prefix : string
+ Prefix to attach to the ID. Usually use obj's non-unique
+ name as prefix.
+
+ assert_on_missing : bool
+ Asserts or not if object doesn't have a registered ID.
+ """
+ prefix = prefix.replace('.', '_')
+ if not hasattr(get_or_create_dot_id, "obj_id_dict"):
+ get_or_create_dot_id.obj_id_dict = {}
+ if obj not in get_or_create_dot_id.obj_id_dict:
+ if assert_on_missing:
+ assert False, 'dot_id ' + str(obj) + ' has not been registered.'
+ else:
+ get_or_create_dot_id.obj_id_dict[obj] = prefix + hex(id(obj))
+ return get_or_create_dot_id.obj_id_dict[obj]
+
+
+def get_port_id(is_input, index):
+ return 'I_' + str(index) if is_input else 'O_' + str(index)
+
+
+def get_itervar_type_info(iter_type):
+ assert iter_type < len(
+ ITERVAR_TYPE_STRING_MAP), 'Unknown IterVar type: ' + str(iter_type)
+ return ITERVAR_TYPE_STRING_MAP[iter_type]
+
+
+def get_itervar_label_color(itervar, iv_type):
+ type_info = get_itervar_type_info(iv_type)
+ return str(itervar.var) + '(' + type_info[0] + ')', type_info[1]
+
+
+def linebrk(s, n):
+ """ Break input string s with <br/> for every n charactors."""
+ result = ''
+ j = 0
+ for i, c in enumerate(s):
+ if j == n and i != len(s) - 1:
+ result = result + '\n'
+ j = 0
+ j = j + 1
+ result = result + c
+ result = html.escape(str(result), quote=True)
+ result = result.replace('\n', '<br/>')
+ return result
+
+
+def create_graph(name="", rankdir='BT'):
+ graph = Digraph(name=name)
+ graph.graph_attr['rankdir'] = rankdir
+ return graph
+
+
+def itervar_label(itervar, index, index_color, label):
+ return '<TR><TD PORT="' + get_or_create_dot_id(itervar, str(
+ itervar.var)) + '" BGCOLOR="' + index_color + '">' + str(
+ index
+ ) + '</TD><TD BGCOLOR="white" PORT="itervar">' + label + '</TD></TR>'
+
+
+def stage_label(stage):
+ return stage.op.name + '<br/>Scope: ' + stage.scope
+
+
+def legend_label():
+ label = '<<TABLE BORDER="0" CELLBORDER="1" CELLSPACING="0" CELLPADDING="4">'
+ for iter_type in ITERVAR_TYPE_STRING_MAP:
+ name, color = ITERVAR_TYPE_STRING_MAP[iter_type]
+ label += '<TR><TD BGCOLOR="' + color + '"></TD>' \
+ + '<TD BGCOLOR="white">' + name + '</TD></TR>'
+ label += '</TABLE>>'
+ return label
+
+
+def legend_dot(g):
+ with g.subgraph(name='cluster_legend') as subgraph:
+ subgraph.attr(label='Legend')
+ label = legend_label()
+ subgraph.node('legend', label, shape='none', margin='0')
+
+
+def dump_graph(dot_string,
+ showsvg=True,
+ dotfilepath='',
+ outputdotstring=False):
+ """Output dot_string in various formats."""
+ if dotfilepath:
+ try:
+ dot_file = open(dotfilepath, "w+")
+ dot_file.write(dot_string)
+ dot_file.close()
+ except IOError:
+ print('Cannot open file: ' + dotfilepath)
+ if showsvg:
+ src = Source(dot_string)
+ display(SVG(src.pipe(format='svg')))
+ if outputdotstring:
+ return dot_string
+ return None
+
+
+def viz_schedule_tree(sch,
+ showsvg=False,
Review comment:
I have disabled TEDD execution and verification and kept TE construction in the test.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] tqchen edited a comment on issue #4651: Tensor
Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
tqchen edited a comment on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-591231104
Sorry about the delay. Thanks @yongfeng-nv @Hzfengsy this PR is now merged!
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on issue #4651: Tensor
Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-572148284
One question about test failure: https://ci.tvm.ai/blue/organizations/jenkins/tvm/detail/PR-4651/2/pipeline
> E ModuleNotFoundError: No module named 'graphviz'
It fails, because tedd depends on graphviz, but tvmai/ci-cpu:v0.54 doesn't have it.
How to fix this issue? Let ci-cpu install graphviz as ci-gpu?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on issue #4651: Tensor
Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-584702130
The PR is clear from building/testing failures, after qualifying with @Hzfengsy's help on tutorial.
I am attaching the three static images used by the tutorial. If there is no more comments for me to address, can someone help uploading them as: https://github.com/dmlc/web-data/raw/master/tvm/tutorial/tedd_itervar_rel.png, https://github.com/dmlc/web-data/raw/master/tvm/tutorial/tedd_dfg.png, and https://github.com/dmlc/web-data/raw/master/tvm/tutorial/tedd_st.png? Thanks
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on issue #4651: Tensor
Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-578845333
> > Thanks @yongfeng-nv . One thing that I think worth considering, as in many viz tools, is the separation of visualization data source specification(in this case the perhaps a dom tree or similar kind) from the visualization(graphviz).
> > We can have a tool that extracts the spec into json, then have a tool to take that spec and visualize it
>
> @tqchen, I see your point. Let me spec the data source spec.
@tqchen, @Hzfengsy, I have posted proposal DOM tree in the RFC thread: https://discuss.tvm.ai/t/visualize-tensor-expression/5174/6. Please leave your comments and suggestions.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-tvm] yongfeng-nv commented on issue #4651: Tensor
Expression Debug Display (TEDD)
Posted by GitBox <gi...@apache.org>.
yongfeng-nv commented on issue #4651: Tensor Expression Debug Display (TEDD)
URL: https://github.com/apache/incubator-tvm/pull/4651#issuecomment-584163670
@tqchen, @Hzfengsy
I have updated this PR with changes listed in the forum:
https://discuss.tvm.ai/t/visualize-tensor-expression/5174/12?u=maplegu
Two questions about the tutorial:
1. As graphviz is not always available in CI. Shall I use static images in the tutorial? If so, my current submission uses image locations such as https://github.com/dmlc/web-data/raw/master/tvm/tutorial/tedd_st.png. I haven't committed any image yet. Is it a good location?
2. I would like to inspect the rendered tutorial. How to view it in my repo?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services