debugging mnist model

.gitignore

@@ -4,4 +4,5 @@ flagged/*
 data_mnist/*
 model/*
 model
-data_mnist
+data_mnist
+slurm*

README.md

@@ -1,12 +1,190 @@
----
-title: MNIST Adversarial
-emoji: 🔢
-colorFrom: gray
-colorTo: indigo
-sdk: gradio
-sdk_version: 3.0.20
-app_file: app.py
-pinned: false
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Git Large File Storage
+
+![CI status][1]
+
+[1]: https://github.com/git-lfs/git-lfs/workflows/CI/badge.svg
+
+[Git LFS](https://git-lfs.github.com) is a command line extension and
+[specification](docs/spec.md) for managing large files with Git.
+
+The client is written in Go, with pre-compiled binaries available for Mac,
+Windows, Linux, and FreeBSD. Check out the [website](http://git-lfs.github.com)
+for an overview of features.
+
+## Getting Started
+
+### Downloading
+
+You can install the Git LFS client in several different ways, depending on your
+setup and preferences.
+
+* **Linux users**. Debian and RPM packages are available from
+  [PackageCloud](https://packagecloud.io/github/git-lfs/install).
+* **macOS users**. [Homebrew](https://brew.sh) bottles are distributed, and can
+  be installed via `brew install git-lfs`.
+* **Windows users**. Git LFS is included in the distribution of
+  [Git for Windows](https://gitforwindows.org/). Alternatively, you can
+  install a recent version of Git LFS from the [Chocolatey](https://chocolatey.org/) package manager.
+* **Binary packages**. In addition, [binary packages](https://github.com/git-lfs/git-lfs/releases) are
+available for Linux, macOS, Windows, and FreeBSD.
+* **Building from source**. [This repository](https://github.com/git-lfs/git-lfs.git) can also be
+built from source using the latest version of [Go](https://golang.org), and the
+available instructions in our
+[Wiki](https://github.com/git-lfs/git-lfs/wiki/Installation#source).
+
+### Installing
+
+#### From binary
+
+The [binary packages](https://github.com/git-lfs/git-lfs/releases) include a script which will:
+
+- Install Git LFS binaries onto the system `$PATH`
+- Run `git lfs install` to
+perform required global configuration changes.
+
+```ShellSession
+$ ./install.sh
+```
+
+#### From source
+
+- Place the `git-lfs` binary on your system’s executable `$PATH` or equivalent.
+- Git LFS requires global configuration changes once per-machine. This can be done by
+running:
+
+```ShellSession
+$ git lfs install
+```
+
+## Example Usage
+
+To begin using Git LFS within a Git repository that is not already configured
+for Git LFS, you can indicate which files you would like Git LFS to manage.
+This can be done by running the following _from within a Git repository_:
+
+```bash
+$ git lfs track "*.psd"
+```
+
+(Where `*.psd` is the pattern of filenames that you wish to track. You can read
+more about this pattern syntax
+[here](https://git-scm.com/docs/gitattributes)).
+
+> *Note:* the quotation marks surrounding the pattern are important to
+> prevent the glob pattern from being expanded by the shell.
+
+After any invocation of `git-lfs-track(1)` or `git-lfs-untrack(1)`, you _must
+commit changes to your `.gitattributes` file_. This can be done by running:
+
+```bash
+$ git add .gitattributes
+$ git commit -m "track *.psd files using Git LFS"
+```
+
+You can now interact with your Git repository as usual, and Git LFS will take
+care of managing your large files. For example, changing a file named `my.psd`
+(tracked above via `*.psd`):
+
+```bash
+$ git add my.psd
+$ git commit -m "add psd"
+```
+
+> _Tip:_ if you have large files already in your repository's history, `git lfs
+> track` will _not_ track them retroactively. To migrate existing large files
+> in your history to use Git LFS, use `git lfs migrate`. For example:
+>
+> ```
+> $ git lfs migrate import --include="*.psd"
+> ```
+>
+> For more information, read [`git-lfs-migrate(1)`](https://github.com/git-lfs/git-lfs/blob/master/docs/man/git-lfs-migrate.1.ronn).
+
+You can confirm that Git LFS is managing your PSD file:
+
+```bash
+$ git lfs ls-files
+3c2f7aedfb * my.psd
+```
+
+Once you've made your commits, push your files to the Git remote:
+
+```bash
+$ git push origin master
+Uploading LFS objects: 100% (1/1), 810 B, 1.2 KB/s
+# ...
+To https://github.com/git-lfs/git-lfs-test
+   67fcf6a..47b2002  master -> master
+```
+
+Note: Git LFS requires at least Git 1.8.2 on Linux or 1.8.5 on macOS.
+
+## Limitations
+
+Git LFS maintains a list of currently known limitations, which you can find and
+edit [here](https://github.com/git-lfs/git-lfs/wiki/Limitations).
+
+## Need Help?
+
+You can get help on specific commands directly:
+
+```bash
+$ git lfs help <subcommand>
+```
+
+The [official documentation](docs) has command references and specifications for
+the tool.
+
+You can always [open an issue](https://github.com/git-lfs/git-lfs/issues), and
+one of the Core Team members will respond to you. Please be sure to include:
+
+1. The output of `git lfs env`, which displays helpful information about your
+   Git repository useful in debugging.
+2. Any failed commands re-run with `GIT_TRACE=1` in the environment, which
+   displays additional information pertaining to why a command crashed.
+
+## Contributing
+
+See [CONTRIBUTING.md](CONTRIBUTING.md) for info on working on Git LFS and
+sending patches. Related projects are listed on the [Implementations wiki
+page](https://github.com/git-lfs/git-lfs/wiki/Implementations).
+
+## Core Team
+
+These are the humans that form the Git LFS core team, which runs the project.
+
+In alphabetical order:
+
+| [@bk2204][bk2204-user] | [@larsxschneider][larsxschneider-user] | [@PastelMobileSuit][PastelMobileSuit-user] |
+|---|---|---|
+| [![][bk2204-img]][bk2204-user] | [![][larsxschneider-img]][larsxschneider-user] | [![][PastelMobileSuit-img]][PastelMobileSuit-user] |
+
+[bk2204-img]: https://avatars1.githubusercontent.com/u/497054?s=100&v=4
+[larsxschneider-img]: https://avatars1.githubusercontent.com/u/477434?s=100&v=4
+[PastelMobileSuit-img]: https://avatars2.githubusercontent.com/u/37254014?s=100&v=4
+[bk2204-user]: https://github.com/bk2204
+[larsxschneider-user]: https://github.com/larsxschneider
+[PastelMobileSuit-user]: https://github.com/PastelMobileSuit
+
+### Alumni
+
+These are the humans that have in the past formed the Git LFS core team, or
+have otherwise contributed a significant amount to the project. Git LFS would
+not be possible without them.
+
+In alphabetical order:
+
+| [@andyneff][andyneff-user] | [@rubyist][rubyist-user] | [@sinbad][sinbad-user] | [@technoweenie][technoweenie-user] | [@ttaylorr][ttaylorr-user] |
+|---|---|---|---|---|
+| [![][andyneff-img]][andyneff-user] | [![][rubyist-img]][rubyist-user] | [![][sinbad-img]][sinbad-user] | [![][technoweenie-img]][technoweenie-user] | [![][ttaylorr-img]][ttaylorr-user] |
+
+[andyneff-img]: https://avatars1.githubusercontent.com/u/7596961?v=3&s=100
+[rubyist-img]: https://avatars1.githubusercontent.com/u/143?v=3&s=100
+[sinbad-img]: https://avatars1.githubusercontent.com/u/142735?v=3&s=100
+[technoweenie-img]: https://avatars3.githubusercontent.com/u/21?v=3&s=100
+[ttaylorr-img]: https://avatars2.githubusercontent.com/u/443245?s=100&v=4
+[andyneff-user]: https://github.com/andyneff
+[sinbad-user]: https://github.com/sinbad
+[rubyist-user]: https://github.com/rubyist
+[technoweenie-user]: https://github.com/technoweenie
+[ttaylorr-user]: https://github.com/ttaylorr

app.py

@@ -20,11 +20,12 @@ n_epochs = 10
 batch_size_train = 128
 batch_size_test = 1000
 learning_rate = 0.01
+adv_learning_rate= 0.001
 momentum = 0.5
 log_interval = 10
 random_seed = 1
 TRAIN_CUTOFF = 10
-TEST_PER_SAMPLE = 1500
+TEST_PER_SAMPLE = 5000
 DASHBOARD_EXPLANATION = DASHBOARD_EXPLANATION.format(TEST_PER_SAMPLE=TEST_PER_SAMPLE)
 WHAT_TO_DO=WHAT_TO_DO.format(num_samples=TRAIN_CUTOFF)
 MODEL_PATH = 'model' 
@@ -162,7 +163,6 @@ TRAIN_TRANSFORM = torchvision.transforms.Compose([
 test_loader = torch.utils.data.DataLoader(MNISTCorrupted(TRAIN_TRANSFORM),
                                             batch_size=batch_size_test, shuffle=False)
 
-
 # Source: https://nextjournal.com/gkoehler/pytorch-mnist
 class MNIST_Model(nn.Module):
     def __init__(self):
@@ -220,6 +220,7 @@ def test():
         acc = acc.item()
         test_metric = '〽Current test metric -> Avg. loss: `{:.4f}`, Accuracy: `{:.0f}%`\n'.format(
         test_loss,acc)
+        print(test_metric)
         return test_metric,acc
 
 
@@ -233,6 +234,34 @@ optimizer = optim.SGD(network.parameters(), lr=learning_rate,
                       momentum=momentum)        
 
 
+
+train_loader = torch.utils.data.DataLoader(
+  torchvision.datasets.MNIST('./files/', train=True, download=True,
+                             transform=TRAIN_TRANSFORM),
+  batch_size=batch_size_train, shuffle=True)
+
+test_iid_loader = torch.utils.data.DataLoader(
+  torchvision.datasets.MNIST('./files/', train=False, download=True,
+                             transform=TRAIN_TRANSFORM),
+  batch_size=batch_size_test, shuffle=True)
+
+model_state_dict = MODEL_WEIGHTS_PATH
+optimizer_state_dict = OPTIMIZER_PATH
+if os.path.exists(model_state_dict) and os.path.exists(optimizer_state_dict):
+    network_state_dict = torch.load(model_state_dict)
+    network.load_state_dict(network_state_dict)
+
+    optimizer_state_dict = torch.load(optimizer_state_dict)
+    optimizer.load_state_dict(optimizer_state_dict)     
+
+# Train model
+#n_epochs=20
+#train(n_epochs,network,optimizer,train_loader)
+#test()
+
+
+
+
 def train_and_test(train_model=True):
 
     if train_model:
@@ -244,6 +273,7 @@ def train_and_test(train_model=True):
     
     test_metric,test_acc = test()
 
+    network.eval()
     if os.path.exists(METRIC_PATH):
         metric_dict = read_json(METRIC_PATH)
         metric_dict['all'] = metric_dict['all']+ [test_acc] if 'all' in metric_dict else [] + [test_acc]
@@ -273,6 +303,7 @@ def train_and_test(train_model=True):
     return test_metric
 
 
+# Update model weights again
 model_state_dict = MODEL_WEIGHTS_PATH
 optimizer_state_dict = OPTIMIZER_PATH
 model_repo.git_pull()
@@ -287,9 +318,14 @@ else:
     # Use best weights  
     BEST_WEIGHTS_MODEL = "best_weights/mnist_model.pth"
     BEST_WEIGHTS_OPTIMIZER = "best_weights/optimizer.pth"
-    torch.save(network.state_dict(), BEST_WEIGHTS_MODEL)
-    torch.save(optimizer.state_dict(), BEST_WEIGHTS_OPTIMIZER)
-    _ = train_and_test(False)
+    
+    network_state_dict = torch.load(BEST_WEIGHTS_MODEL)
+    network.load_state_dict(network_state_dict)
+
+    optimizer_state_dict = torch.load(BEST_WEIGHTS_OPTIMIZER)
+    optimizer.load_state_dict(optimizer_state_dict)        
+    if not os.path.exists(METRIC_PATH):
+        _ = train_and_test(False)
                    
 
 def image_classifier(inp):
@@ -305,20 +341,23 @@ def image_classifier(inp):
     model_repo.git_pull()
     model_state_dict = MODEL_WEIGHTS_PATH
     optimizer_state_dict = OPTIMIZER_PATH
+    which_weights=''
 
     if os.path.exists(model_state_dict) and os.path.exists(optimizer_state_dict):
+        which_weights = "Using weights from model repo"
         network_state_dict = torch.load(model_state_dict)
         network.load_state_dict(network_state_dict)
         optimizer_state_dict = torch.load(optimizer_state_dict)
         optimizer.load_state_dict(optimizer_state_dict)
     else:
-        # Use best weights  
+        # Use best weights 
+        which_weights = "Using default best weights"
         BEST_WEIGHTS_MODEL = "best_weights/mnist_model.pth"
         BEST_WEIGHTS_OPTIMIZER = "best_weights/optimizer.pth"   
         network.load_state_dict(torch.load(BEST_WEIGHTS_MODEL))
         optimizer.load_state_dict(torch.load(BEST_WEIGHTS_OPTIMIZER)) 
-
-    input_image = torchvision.transforms.ToTensor()(inp).unsqueeze(0)
+    network.eval()
+    input_image = TRAIN_TRANSFORM(inp).unsqueeze(0)
     with torch.no_grad():
 
         prediction = torch.nn.functional.softmax(network(input_image)[0], dim=0)
@@ -512,7 +551,7 @@ def main():
 
 
 
-    block.launch()  
+    block.launch(share=True)  
         
      
 

best_weights/mnist_model.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ba8d282674beb300db53069e4972cfed358f8c7c627cf449215e44b365fcdc54
+oid sha256:148112958ca9545938f0660cec604ac4c7f52dca3523091e1e8e4e6a26e1ebc7
 size 89871

best_weights/optimizer.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fe255c0ca501d01ae3c2083ea760ea95759fcfe9075e39fba299c57a9907bf1b
-size 623
+oid sha256:aac1c136737d50c2665563392a5b220396398cf1e2a2049dbefd7dc95473f5a5
+size 89807

data_mnist

@@ -1 +1 @@
-Subproject commit ed62a26e764902f519ff43df850842e07dfe2cc0
+Subproject commit c26d5d2bfb2623f637df6101899ad1f334b05de3

requirements.txt

@@ -1,3 +1,5 @@
 torch
 torchvision
-matplotlib
+matplotlib
+gradio
+huggingface_hub