-
I'd like to be able to override a trained model's batch size more easily during model conversions. For, example, the Squeezenet1.1 deploy model at [https://github.com/DeepScale/SqueezeNet/tree/maste…
-
the inception v2 model has an initial depthwise convolution with a multiplier of 8. Reading your spec, it wasn't clear to me how to do this conversion, but I did find your answer in another post here…
-
I noticed that the output from my current download of nnef tools, using the caffe-to-nnef conversion with the python, results in a 512 byte header, in other words zero padding out to 0x200, while prev…
-
Let me reference @ajtulloch 's [comment](https://github.com/dmlc/tvm/pull/2116#issuecomment-444694200) about quantization workflow firstly:
> 1. Implement a model in a standard ML framework, genera…
-
This page has a several tar'd up training models and eval models for tensorflow in protobuf, tflite and pbtext forms. They also have saved checkpoint data. These seem closer to what was used for c…
-
I see stated in the tensorflow docs that TensorFlow's implementation of padding comes up with different values than Caffe's implementation. Does NNEF comply with the source model's padding values, …
-
In the function `write_tensor` from [`nnef.py`](https://github.com/KhronosGroup/NNEF-Tools/blob/beb6842c6f952eb981571263bbda783774982716/parser/cpp/python/nnef.py#L336) it seems like file header have …
-
I use the last version(6825938) converter(au-zone) from mobilenet[1] to nnef,
But the depthwise operator(tf.nn.depthwise_conv2d_native) is converted to conv(NNEF),
The result should be planewise_con…
-
If you attempt to slice up the resnet graph for pipelined processing, it appears to me that implies some type of per layer identity operations for the bypass loops in the graph.
I'm wondering if t…
-
This can host comments and discussion for now.
Some additional things that would be nice to have but need some thought:
- [x] ~Saving/Loading models~
- [x] ~A Model class~
- [ ] A Trainer that h…