README.md

docker-torch-rnn

Docker images for using torch-rnn (https://github.com/jcjohnson/torch-rnn)

Available tags

• crisbal/torch-rnn:base
  • Based on ubuntu:14.04
  • Allows usage of torch-rnn in CPU mode
• lordalfred/docker-torch-rnn:cuda7.5
  • Based on nvidia/cuda:7.5-devel-ubuntu14.04
  • Allows usage of torch-rnn in GPU mode (Cuda 7.5 support)
  • Only run with nvidia-docker https://github.com/NVIDIA/nvidia-docker
• lordalfred/docker-torch-rnn:cuda8.0
  • Based on nvidia/cuda:8.0-devel-ubuntu16.04
  • Allows usage of torch-rnn in GPU mode (Cuda 8.0 support)
  • Only run with nvidia-docker https://github.com/NVIDIA/nvidia-docker
• lordalfred/docker-torch-rnn:cuda9.1
  • Based on nvidia/cuda:9.1-devel-ubuntu16.04
  • Allows usage of torch-rnn in GPU mode (Cuda 9.1 support)
  • Only run with nvidia-docker https://github.com/NVIDIA/nvidia-docker
• lordalfred/docker-torch-rnn:cuda9.2
  • Based on nvidia/cuda:9.2-devel-ubuntu16.04
  • Allows usage of torch-rnn in GPU mode (Cuda 9.2 support)
  • Only run with nvidia-docker https://github.com/NVIDIA/nvidia-docker
• lordalfred/docker-torch-rnn:cuda10.0 (requires nvidia-docker v2)
  • Based on nvidia/cuda:10.0-devel-ubuntu16.04
  • Allows usage of torch-rnn in GPU mode (Cuda 10.0 support)
  • Only run with nvidia-docker https://github.com/NVIDIA/nvidia-docker
• lordalfred/docker-torch-rnn:10.0-ubuntu18.04 (requires nvidia-docker v2)
  • Based on nvidia/cuda:10.0-devel-ubuntu18.04
  • Allows usage of torch-rnn in GPU mode (Cuda 10.0 support)
  • Only run with nvidia-docker https://github.com/NVIDIA/nvidia-docker

How to

More details here: https://github.com/jcjohnson/torch-rnn#usage

CPU Only

1. Start bash in the container

   • docker run --rm -ti crisbal/torch-rnn:base bash

2. Preprocess the sample data

   python scripts/preprocess.py \
   --input_txt data/tiny-shakespeare.txt \
   --output_h5 data/tiny-shakespeare.h5 \
   --output_json data/tiny-shakespeare.json
   

3. Train

   th train.lua \
   -input_h5 data/tiny-shakespeare.h5 \
   -input_json data/tiny-shakespeare.json \
   -gpu -1
   

4. Sample

   • th sample.lua -checkpoint cv/checkpoint_10000.t7 -length 2000 -gpu -1

CUDA

1. Install nvidia-docker

   • https://github.com/NVIDIA/nvidia-docker

2. Start bash in the container

   • nvidia-docker run --rm -ti lordalfred/docker-torch-rnn:cuda10.0 bash

3. Preprocess the sample data

   python scripts/preprocess.py \
   --input_txt data/tiny-shakespeare.txt \
   --output_h5 data/tiny-shakespeare.h5 \
   --output_json data/tiny-shakespeare.json
   

4. Train

   th train.lua \
   -input_h5 data/tiny-shakespeare.h5 \
   -input_json data/tiny-shakespeare.json
   

5. Sample

   • th sample.lua -checkpoint cv/checkpoint_10000.t7 -length 2000