The code in models/regularized_gan.py (here) defines the tensorflow model for training a GAN network to generate mnist data.

The if network_type == "mnist": line strongly implies that the model would look different for other tasks.

Can this same model be used for the other tasks described in the paper, such as 3d faces or CelebA? (Assuming we change the input codes, of course.)

Thanks! :)

The README says:

Running Experiment

We provide the source code to run the MNIST example:

python launchers/run_mnist_exp.py

You can launch TensorBoard to view the generated images:

tensorboard --logdir logs/mnist

But running that experiment produces only a single log in logs, but no checkpoints in ckt and no logs after the model has trained. It's disappointing because I ran the experiment over a day and then it finished, and now all that work is gone.

root@cbe2bb86132a:/InfoGAN# find ckt
ckt
ckt/mnist
ckt/mnist/mnist_2016_08_15_03_18_12

root@cbe2bb86132a:/InfoGAN# find logs
logs
logs/mnist
logs/mnist/mnist_2016_08_15_03_18_12
logs/mnist/mnist_2016_08_15_03_18_12/events.out.tfevents.1471231106.cbe2bb86132a

Is there some flag setting I need to set to get useful output? If so, can that be the default? Thanks!

W tensorflow/core/common_runtime/bfc_allocator.cc:271] Ran out of memory trying to allocate 24.50MiB.  See logs for memory state.
W tensorflow/core/framework/op_kernel.cc:909] Resource exhausted: OOM when allocating tensor with shape[6272,1024]

Looks like two devices are being created on the same device [last 2 lines]?

> python launchers/run_mnist_exp.py
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcublas.so locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcudnn.so locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcufft.so locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcuda.so locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcurand.so locally
Extracting MNIST/train-images-idx3-ubyte.gz
Extracting MNIST/train-labels-idx1-ubyte.gz
Extracting MNIST/t10k-images-idx3-ubyte.gz
Extracting MNIST/t10k-labels-idx1-ubyte.gz
I tensorflow/core/common_runtime/gpu/gpu_init.cc:102] Found device 0 with properties:
name: GeForce GTX TITAN X
major: 5 minor: 2 memoryClockRate (GHz) 1.076
pciBusID 0000:06:00.0
Total memory: 12.00GiB
Free memory: 496.40MiB
I tensorflow/core/common_runtime/gpu/gpu_init.cc:126] DMA: 0
I tensorflow/core/common_runtime/gpu/gpu_init.cc:136] 0:   Y
I tensorflow/core/common_runtime/gpu/gpu_device.cc:806] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX TITAN X, pci bus id: 0000:06:00.0)
I tensorflow/core/common_runtime/gpu/gpu_device.cc:806] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX TITAN X, pci bus id: 0000:06:00.0)

Running tf0.9

2022-02-10T10:10:25.079-0800 [TRACE] plugin.terraform-provider-kubernetes_v2.8.0_x5: Served request: tf_req_id=05f706a8-2360-b657-8ac1-faf1085ba708 tf_resource_type=kubernetes_manifest EXTRA_VALUE_AT_END=<nil> tf_proto_version=5.2 tf_provider_addr= tf_rpc=PlanResourceChange @caller=/opt/teamcity-agent/work/5d79fe75d4460a2f/src/github.com/hashicorp/terraform-provider-kubernetes/vendor/github.com/hashicorp/terraform-plugin-go/internal/logging/protocol.go:21 @module=sdk.proto timestamp=2022-02-10T10:10:25.078-0800
2022/02/10 10:10:25 [TRACE] EvalWriteState: writing current state object for module.victoriametrics-cluster-stack.kubernetes_manifest.vmselect-statefulset[0]
2022/02/10 10:10:25 [TRACE] EvalWriteDiff: recorded NoOp change for module.victoriametrics-cluster-stack.kubernetes_manifest.vmselect-statefulset[0]
2022/02/10 10:10:25 [TRACE] vertex "module.victoriametrics-cluster-stack.kubernetes_manifest.vmselect-statefulset[0]": visit complete
2022/02/10 10:10:25 [TRACE] vertex "module.victoriametrics-cluster-stack.kubernetes_manifest.vmselect-statefulset": dynamic subgraph completed successfully
2022/02/10 10:10:25 [TRACE] vertex "module.victoriametrics-cluster-stack.kubernetes_manifest.vmselect-statefulset": visit complete
2022/02/10 10:10:25 [TRACE] vertex "module.victoriametrics-cluster-stack.kubernetes_manifest.vmselect-statefulset (expand)": dynamic subgraph completed successfully
2022/02/10 10:10:25 [TRACE] vertex "module.victoriametrics-cluster-stack.kubernetes_manifest.vmselect-statefulset (expand)": visit complete
2022/02/10 10:10:25 [TRACE] dag/walk: upstream of "module.victoriametrics-cluster-stack.provider[\"registry.terraform.io/hashicorp/kubernetes\"] (close)" errored, so skipping
2022/02/10 10:10:25 [TRACE] dag/walk: upstream of "module.victoriametrics-cluster-stack (close)" errored, so skipping
2022/02/10 10:10:25 [TRACE] dag/walk: upstream of "meta.count-boundary (EachMode fixup)" errored, so skipping
2022/02/10 10:10:25 [TRACE] dag/walk: upstream of "root" errored, so skipping
2022/02/10 10:10:25 [INFO] backend/local: plan operation completed

Snyk has created this PR to fix one or more vulnerable packages in the `pip` dependencies of this project.

Changes included in this PR

• Changes to the following files to upgrade the vulnerable dependencies to a fixed version:
  • requirements.txt

Vulnerabilities that will be fixed

By pinning:

Severity | Priority Score (*) | Issue | Upgrade | Breaking Change | Exploit Maturity :-------------------------:|-------------------------|:-------------------------|:-------------------------|:-------------------------|:------------------------- | 661/1000
Why? Recently disclosed, Has a fix available, CVSS 7.5 | Denial of Service (DoS)
SNYK-PYTHON-PYGMENTS-1088505 | pygments:
2.5.2 -> 2.7.4
| No | No Known Exploit

(*) Note that the real score may have changed since the PR was raised.

Some vulnerabilities couldn't be fully fixed and so Snyk will still find them when the project is tested again. This may be because the vulnerability existed within more than one direct dependency, but not all of the effected dependencies could be upgraded.

Check the changes in this PR to ensure they won't cause issues with your project.

Note: You are seeing this because you or someone else with access to this repository has authorized Snyk to open fix PRs.

For more information: 🧐 View latest project report

🛠 Adjust project settings

📚 Read more about Snyk's upgrade and patch logic

Hello, I want to experiment with the "faces" dataset, so I wonder how you gain your faces dataset? I also download the MATLAB code about the "Basel face model", but I didn't find the parameters about the light, pose and the elevation.

The Docker container is mapping port 8888 from guest to host but Tensorboard starts on port 6006. Add port option to Tensorboard to start on the specified port. Add PYTHONPATH environment variable to python command in Docker section to solve issue with conflicting imports.

Replace link to gcr.io and link instead to the same image on Docker Hub as it doesn't work anymore.

See link to tensorflow/tensorflow:r0.9rc0-devel.

Fixes https://github.com/openai/InfoGAN/issues/30.

when I run

docker pull gcr.io/tensorflow/tensorflow:r0.9rc0-devel

Occur this error below.

Error response from daemon: manifest for gcr.io/tensorflow/tensorflow:r0.9rc0-devel not found

docker image not found.

Hi,

I read the paper of InfoGAN (https://arxiv.org/abs/1606.03657) and I found that it has the objective function for Discriminator (D), Generator (G), and mutual information network (Q) in equation (6).

So, in the implementation, D is to maximize the part which originally for GAN, G is to minimize its own part for GAN minus the mutual information times a tuning lambda, and Q is to maximize the mutual information as it uses a lower bound for mutual information.

However, in this InfoGAN official code, I actually found in InfoGAN/infogan/algos/infogan_trainer.py that:

discriminator_loss -= self.info_reg_coeff * disc_mi_est
generator_loss -= self.info_reg_coeff * disc_mi_est

So both the discriminator loss and the generator loss incorporate the mutual information lower bound. I found this is not consistent with the paper.

First, the discriminator loss does not have the mutual information lower bound in the formulation. Although the paper says it constructed Q network from the last hidden layer of D, training the D network should not incorporate the mutual information lower bound value.

Second, in equation (6) or any other GANs, G and D should always have opposite signs (+ or -) for the same term. But the paired code above used mutual information both to be the negative sign for D and G to minimize. This does not make sense and it means the equation (6) in paper is not correct and should be changed.

Could you help me with this?