[DARKNET]RNN Support for darknet #1443
Conversation
|
@PariksheetPinjari909 , @srkreddy1238 @Huyuwei @masahi |
|
|
||
| target = 'llvm' | ||
| shape_dict = {'data': data.shape} | ||
| #with nnvm.compiler.build_config(opt_level=2): |
There was a problem hiding this comment.
remove the commented line if not required.
| shape_dict = {'data': data.shape} | ||
| #with nnvm.compiler.build_config(opt_level=2): | ||
| graph, library, params = nnvm.compiler.build(sym, target, shape_dict, dtype, params=params) | ||
| ###################################################################### |
There was a problem hiding this comment.
Correct the comment styles.
| ###################################################################### | ||
| # Prepare cfg and weights file | ||
| # Pretrained model available | ||
| # -------------------------------------------------------------------- |
There was a problem hiding this comment.
Doc style corrections.
|
|
||
| CFG_NAME = MODEL_NAME + '.cfg' | ||
| WEIGHTS_NAME = MODEL_NAME + '.weights' | ||
| CFG_URL = 'https://github.com/siju-samuel/darknet/blob/master/cfg/' + \ |
There was a problem hiding this comment.
may use dmlc/web-data instead
|
|
||
| ###################################################################### | ||
| # Compile the model on NNVM | ||
| # -------------------------------------------------------------------- |
There was a problem hiding this comment.
doc style corrections
tqchen
added
status: review in progress
status: need update
|
@srkreddy1238 Thanks for the comments. All comments are addressed. Please check again. TIA |
| CFG_URL = REPO_URL + 'cfg/' + CFG_NAME + '?raw=true' | ||
| WEIGHTS_URL = REPO_URL + 'weights/' + WEIGHTS_NAME + '?raw=true' | ||
|
|
||
| def _dl_progress(count, block_size, total_size): |
There was a problem hiding this comment.
Tutorial scripts gets exported to docs.tvm.ai (Ex: https://docs.tvm.ai/tutorials/nnvm/get_started.html) .
Hence suggest to simplify the example by using existing methods.
Here we could use glueon download methods instead of rewriting.
| with nnvm.compiler.build_config(opt_level=2): | ||
| graph, lib, params = nnvm.compiler.build(sym, target, shape_dict, dtype_dict, params) | ||
|
|
||
| # Save the json |
There was a problem hiding this comment.
Saving lib already explained in get_started.py. Can leave a doc pointer instead.
|
@siju-samuel with above comments LGTM. |
|
@srkreddy1238 Thanks for the review. |
tqchen
added
status: accepted
and removed
status: need update
|
Thanks @siju-samuel @srkreddy1238 @PariksheetPinjari909 , this is now merged |
RNN Support for DARKNET frontend is added.
This PR consists of
Thanks for contributing to TVM! Please refer to guideline https://docs.tvm.ai/contribute/ for useful information and tips. After the pull request is submitted, please request code reviews from others in the community.