Remove unused files, update transformers to 4.38.1 (#18)

* Remove unused files, update transformers to 4.38.1 and TGI to 1.4.2

README.md

@@ -11,7 +11,7 @@ The [`examples`](./examples) directory contains examples for using the container
 _Note: we added the latest TGI version as an example into the repository, which can be build with._
 
 ```bash
-docker build -t us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-generation-inference-gpu.1.3.4 -f containers/tgi/gpu/1.3.4/Dockerfile .
+docker build -t us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-generation-inference-gpu.1.4.2 -f containers/tgi/gpu/1.4.2/Dockerfile .
 ```
 
 ### Mistral 7B test
@@ -29,7 +29,7 @@ docker run --gpus all -ti -p 8080:80 \
   -e NUM_SHARD=$num_shard \
   -e MAX_INPUT_LENGTH=$max_input_length \
   -e MAX_TOTAL_TOKENS=$max_total_tokens \
-  us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-generation-inference-gpu.1.3.4  
+  us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-generation-inference-gpu.1.4.2  
 ```
 
 Send request:
@@ -41,10 +41,10 @@ curl 127.0.0.1:8080/generate \
     -H 'Content-Type: application/json'
 ```
 
-### Golden Gate Test
+### Gemma Test
 
 ```bash
-model=gg-hf/golden-gate-7b
+model=google/gemma-7b
 num_shard=1
 max_input_length=512
 max_total_tokens=1024
@@ -58,7 +58,7 @@ docker run --gpus all -ti -p 8080:80 \
   -e MAX_TOTAL_TOKENS=$max_total_tokens \
   -e MAX_BATCH_PREFILL_TOKENS=$max_batch_prefill_tokens \
   -e HUGGING_FACE_HUB_TOKEN=$token \
-  us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-generation-inference-gpu.1.3.4
+  us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-generation-inference-gpu.1.4.2
 ```
 
 Send request:
@@ -70,7 +70,7 @@ curl 127.0.0.1:8080/generate \
     -H 'Content-Type: application/json'
 ```
 
-For a Vertex AI example checkout [Deploy Golden Gate on Vertex AI](./examples//vertex-ai/deploy-golden-gate-on-vertex-ai.ipynb)
+For a Vertex AI example checkout [Deploy Gemma on Vertex AI](./examples/vertex-ai/notebooks/deploy-gemma-on-vertex-ai.ipynb)
 
 
 ## Configurations
@@ -87,13 +87,13 @@ After the containers are built, you can run the tests in the `tests` directory t
 
 | Container Tag                                                                 | Framework | Type      | Accelerator |
 | ----------------------------------------------------------------------------- | --------- | --------- | ----------- |
-| [pytorch-training-gpu.2.1.transformers.4.37.2.py310](./containers/pytorch/training/gpu/2.1/transformers/4.37.2/py310/Dockerfile)  | Pytorch   | training  | GPU         |
-| [text-generation-inference-gpu.1.3.4](https://github.com/huggingface/Google-Cloud-Containers/blob/main/containers/tgi/gpu/1.3.4/Dockerfile)                 | -         | inference | GPU         |
+| [pytorch-training-gpu.2.1.transformers.4.38.1.py310](./containers/pytorch/training/gpu/2.1/transformers/4.38.1/py310/Dockerfile)  | Pytorch   | training  | GPU         |
+| [text-generation-inference-gpu.1.4.2](./containers/tgi/gpu/1.4.2/Dockerfile)                 | -         | inference | GPU         |
 
 ## Directory Structure
 
-The container files are organized in a nested folder structure based on the container tag. For example, the Dockerfile for the container with the tag `pytorch-training-gpu.2.0.transformers.4.35.0.py310` is located at `pytorch/training/gpu/2.0/transformers/4.35.0/py310/Dockerfile`.
+The container files are organized in a nested folder structure based on the container tag. For example, the Dockerfile for the container with the tag `pytorch-training-gpu.2.1.transformers.4.38.1.py310` is located at `pytorch/training/gpu/2.1/transformers/4.38.1/py310/Dockerfile`.
 
 ## Updates
 
-When we update the transformers version, we add a new folder in the `transformers` directory. For example, if we update the transformers version to 4.36.0, we would add a new folder at `pytorch/training/gpu/2.0/transformers/4.36.0`.
+When we update the transformers version, we add a new folder in the `transformers` directory. For example, if we update the transformers version to 4.39.0, we would add a new folder at `pytorch/training/gpu/2.0/transformers/4.39.0`.

containers/jax/training/gpu/0.4/transformers/4.37.2/py310/Dockerfile → containers/jax/training/gpu/0.4/transformers/4.38.1/py310/Dockerfile

@@ -7,7 +7,7 @@ ARG DEBIAN_FRONTEND=noninteractive
 # Versions
 ARG CUDA="12"
 ARG CUDNN="89"
-ARG TRANSFORMERS='4.37.2'
+ARG TRANSFORMERS='4.38.1'
 ARG DIFFUSERS='0.26.1'
 ARG DATASETS='2.16.1'
 # jax and flax compatible with transformers ["jax>=0.4.1,<=0.4.13", "jaxlib>=0.4.1,<=0.4.13", "flax>=0.4.1,<=0.7.0"] as mentioned in setup.py

containers/pytorch/training/gpu/2.1/transformers/4.37.2/py310/Dockerfile → containers/pytorch/training/gpu/2.1/transformers/4.38.1/py310/Dockerfile

@@ -7,7 +7,7 @@ ARG DEBIAN_FRONTEND=noninteractive
 
 # Versions
 ARG FLASH_ATTN='2.5.2'
-ARG TRANSFORMERS='4.37.2'
+ARG TRANSFORMERS='4.38.1'
 ARG DIFFUSERS='0.26.1'
 ARG PEFT='0.8.2'
 ARG TRL='0.7.10'

containers/pytorch/training/gpu/2.1/transformers/4.38.0.dev0/README.md → containers/pytorch/training/gpu/2.1/transformers/4.38.1/py310/README.md

@@ -1,12 +1,11 @@
-# Fine-tune Gemma-2B on Vertex AI WorkBench
+# Fine-tune Gemma-7B on Vertex AI WorkBench
 
-This file contains step by step instructions on how to build a docker image and then run it to test the Gemma-2B model using the 
+This file contains step by step instructions on how to build a docker image and then run it to test the Gemma-7B model using the 
 [gemma-finetuning-clm-lora-sft.ipynb](https://github.com/huggingface/Google-Cloud-Containers/blob/main/examples/vertex-ai/gemma-finetuning-clm-lora-sft.ipynb) Notebook on a Vertex AI WorkBench instance and on your local machine.
 
 ## Pre-requisites:
-1. Access to [gg-hf](https://huggingface.co/gg-hf) on Hugging Face Hub in order to download the model and the tokenizer.
+1. Access Terms and Conditions on [Hugging Face Hub](https://huggingface.co/google/gemma-7b) in order to download the model and the tokenizer.
 2. Access to [Google-Cloud-Containers](https://github.com/huggingface/Google-Cloud-Containers) GitHub repository in order to access the docker file.
-3. Access to [new-model-addition-golden-gate](https://github.com/huggingface/new-model-addition-golden-gate/) GitHub repository in order to use transformer library with the gg-hf model integrated into it.
 
 
 We use the [gemma-finetuning-clm-lora-sft.ipynb](https://github.com/huggingface/Google-Cloud-Containers/blob/main/examples/vertex-ai/gemma-finetuning-clm-lora-sft.ipynb) Notebook to test the model.
@@ -18,17 +17,10 @@ Use the following command to build the docker image. Make sure to replace the va
 ```bash
 git clone https://github.com/huggingface/Google-Cloud-Containers
 cd Google-Cloud-Containers
-export GITHUB_TOKEN=your-github-token
-docker build --secret id=GITHUB_TOKEN,env=GITHUB_TOKEN -t pytorch-training-gpu.2.1.transformers.4.38.0.dev0.py310 -f containers/pytorch/training/gpu/2.1/transformers/4.38.0.dev0/py310/Dockerfile .
+docker build -t pytorch-training-gpu.2.1.transformers.4.38.1.py310 -f containers/pytorch/training/gpu/2.1/transformers/4.38.1/py310/Dockerfile .
 ```
 
-For setting the value of `GITHUB_TOKEN` please follow the detailed instructions mentioned in the following links: 
-- [Creating a fine-grained personal access token](https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens#creating-a-fine-grained-personal-access-token)
-
-- [Creating a personal access token (classic)](https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens#creating-a-personal-access-token-classic)
-
-
-## Using Vertex AI WorkBench Instance to fine-tune the Gemma-2B model
+## Using Vertex AI WorkBench Instance to fine-tune the Gemma-7B model
 
 It consists of the following steps:
 1. Push the docker image to the Google Cloud Artifact registry.
@@ -56,12 +48,12 @@ Now, you can push the image to the Google Cloud Artifact registry using the foll
 REGION="us-central1"
 DOCKER_ARTIFACT_REPO="deep-learning-images"
 PROJECT_ID="gcp-project-id"
-BASE_IMAGE="pytorch-training-gpu.2.1.transformers.4.38.0.dev0.py310"
+BASE_IMAGE="pytorch-training-gpu.2.1.transformers.4.38.1.py310"
 FRAMEWORK="pytorch"
 TYPE="training"
 ACCELERATOR="gpu"
 FRAMEWORK_VERSION="2.1"
-TRANSFORMERS_VERISON="4.38.0.dev0"
+TRANSFORMERS_VERISON="4.38.1"
 PYTHON_VERSION="py310"
 
 SERVING_CONTAINER_IMAGE_URI="${REGION}-docker.pkg.dev/${PROJECT_ID}/${DOCKER_ARTIFACT_REPO}/huggingface-${FRAMEWORK}-${TYPE}-${ACCELERATOR}.${FRAMEWORK_VERSION}.transformers.${TRANSFORMERS_VERISON}.${PYTHON_VERSION}:latest"
@@ -109,7 +101,7 @@ We will use the Google Cloud CLI to create a Vertex AI WorkBench instance from a
 
 ```bash
 gcloud notebooks instances create example-instance-1 \
-    --container-repository=us-central1-docker.pkg.dev/gcp-project-id/deep-learning-images/huggingface-pytorch-training-gpu.2.1.transformers.4.38.0.dev0.py310 \
+    --container-repository=us-central1-docker.pkg.dev/gcp-project-id/deep-learning-images/huggingface-pytorch-training-gpu.2.1.transformers.4.38.1.py310 \
     --container-tag=latest \
     --machine-type=n1-standard-4 \
     --location=us-central1-c \
@@ -137,7 +129,7 @@ Then, you can access the [gemma-finetuning-clm-lora-sft.ipynb](https://github.co
 Make sure you have the [gemma-finetuning-clm-lora-sft.ipynb](https://github.com/huggingface/Google-Cloud-Containers/blob/main/examples/vertex-ai/gemma-finetuning-clm-lora-sft.ipynb) Notebook on your local machine. As we are mounting the current directory to the docker container.
 
 ```bash
-docker run -it --gpus all -p 8080:8080 -v $(pwd):/workspace pytorch-training-gpu.2.1.transformers.4.38.0.dev0.py310
+docker run -it --gpus all -p 8080:8080 -v $(pwd):/workspace pytorch-training-gpu.2.1.transformers.4.38.1.py310
 ```
 
 Inside the docker container, you can run the following command to start the jupyter notebook: