Update README.md

README.md

@@ -9,7 +9,7 @@ including scalar, image, histogram, audio, text, graph and embedding.
 
 see [demo](http:35.197.26.245:6006) (result of `demo.py` and some images generated by BEGAN)
 
-## install
+## Install
 
 `#tested on anaconda2/anaconda3, tensorflow 1.2.1, pytorch 0.1.12, torchvision 0.1.8`
 
@@ -19,7 +19,10 @@ see [demo](http:35.197.26.245:6006) (result of `demo.py` and some images generat
 
 As this time (v0.1.12), to use the graph drawing feature, you need to build pytorch from source.
 
-## usage
+## API
+http://tensorboard-pytorch.readthedocs.io/en/latest/tensorboard.html
+
+## Usage
 ```python
 import torch
 import torchvision.utils as vutils
@@ -52,6 +55,7 @@ writer.close()
 
 `tensorboard --logdir runs`  
 
+## Screenshots
 scalar</br>
 <img src="screenshots/scalar.png" height="200">
 
@@ -61,73 +65,24 @@ image</br>
 audio</br>
 <img src="screenshots/audio.png" height="100">
 
-distribution</br>
-<img src="screenshots/distribution.png" height="200">
-
 histogram</br>
 <img src="screenshots/histogram.png" height="400">
 
 text</br>
 <img src="screenshots/text.png" height="200">
-## tweaks
+
+## Tweaks
 To show more images in tensorboard's image tab, just
 modify the hardcoded `event_accumulator` in 
 `~/anaconda3/lib/python3.6/site-packages/tensorflow/tensorboard/backend/application.py`
 as you wish.
 
-## API
-
-`class SummaryWriter(str log_dir)`
-###### parameter:
-* `log_dir` (*str*) - the location of the log folder.
-
-> `add_scalar(str tag, float value, int global_step=None)`
->> ###### parameter:
-> * `value` (*float*) - the value to keep.
- 
-> `add_image(str tag, torch.Tensor t, int global_step=None)`
->> ###### parameter:
-> * `t` (*torch.Tensor*) - torch tensor of size (3,H,W). I suggest use `torchvision.utils.make_grid()` to prepare it. 
-
-> `add_audio(str tag, torch.Tensor t, int global_step=None)`
->> ###### parameter:
-> * `t` (*torch.Tensor*) - one dimensional torch tensor. The value should between [-1, 1]. The sample rate is currently fixed at 44100 KHz.
-
->`add_histogram(str tag, numpy.ndarray values, int global_step=None, bins='tensorflow')`
->> ###### parameter:
-> * `values` (*numpy.ndarray*) - one dimensional numpy array.
-> * `bins` (*str*) - one of {'tensorflow', [numpy_arguments](https://docs.scipy.org/doc/numpy/reference/generated/numpy.histogram.html)}, determines how the bins are made.
-
->`add_text(str tag, str text_to_save, int  global_step=None)`
->> ###### parameter:
-> * `text_to_save ` (*str*) - the string to keep.
-
->`add_graph(torch.nn.Module model, torch.autograd.Variable lastnode)`
->> ###### parameter:
-> * `model ` (*torch.nn.Module*) - the model to draw.
-> * `lastnode ` (*torch.autograd.Variable*) - the output variable of the model.
-
->> see `demo_graph.py`
-
-`add_embedding(torch.Tensor t, str save_path, list metadata=None, torch.Tensor label_img=None)`
-> ###### parameter:
- * `t ` (*torch.Tensor*) - A two dimensional tensor. Each row corresponds to one feature vector.
- * `save_path` (*str*) - save path
- * `metadata` (*list*) - a list of label composed of number or string. `len(metadata)` should equal the number of rows of `t`
- * `label_img` (*torch.Tensor*) - 4-D tensor (NCHW). `label_img.size(0)` should equal the number of rows of `t`
-
-> see `demo_embedding.py`
-
-> ### Common parameters:
-> * `tag` (*str*) - values with same tag group together.
-> * `global_step` (*int*) - logs the training step on which the value is saved.
-
 
 ## TODO
 audio sample rate
 
 
-## reference:
+## Reference:
 
 https://github.com/TeamHG-Memex/tensorboard_logger