Update README.md

README.md

@@ -1,76 +1,4 @@
 # HW4 - Efficient Training and Inference
 
-## Setting up
+## Author: Mira Xiao(minxiao)
 
-### AWS
-This assignment is best completed on AWS. See our documentation here:
-https://cmu.app.box.com/file/1999311311551?s=mhbn89gtmlgqo8eg5qgb35exlds6dslj
-
-### Python environment
-1. Install conda by entering these commands into your command line and following the instructions: 
-```bash
-wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
-bash Miniconda3-latest-Linux-x86_64.sh
-source ~/.bashrc
-```
-
-2. Create conda environment:
-   If you run into error like `UnavailableInvalidChannel: HTTP 403 FORBIDDEN for channel <some channel>` on your EC2 instance, you can solve it by running `conda config --remove channels <some channel>`, and make sure you have the default channel by running `conda config --add channels defaults`.
-```bash
-conda create -n cmu-llms-hw4 python=3.11
-conda activate cmu-llms-hw4
-conda install pytorch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1 -c pytorch
-pip install -r requirements.txt
-pip install flash-attn
-pip install wandb
-pip install matplotlib
-pip install seaborn
-pip install vllm
-# Note: huggingface CLI is already installed via requirements.txt
-```
-3. Run `wandb login` to finish setting up weights & biases for experiment tracking (you will need to have a [weights & biases account](https://wandb.ai/login)).  
-
-4. Run `huggingface-cli login` to allow downloading and uploading to Huggingface hub.  
-
-Note: We pin `huggingface_hub[cli]==0.25.2` in `requirements.txt` for compatibility with `transformers==4.45.0`. Avoid upgrading `huggingface-hub` to 1.x, which is incompatible with this version of Transformers and can cause dependency conflicts or a missing `huggingface-cli`.
-
-5. Run `./get_initial_model.sh` to download the model starting point.  
-
-## Contents
-
-This repo contains a simple huggingface-based pre-training script, supporting two datasets:
-
-- two splits of wikitext, with 50M tokens each. Both splits are pre-tokenized for your convinience, one set with sequences of 512 tokens, and the other 2048.
-- minipile, with 1.4B tokens, pre-tokenized with sequences of 2048 tokens.
-
-## Pre-training
-
-The folder ```scripts``` contains access points to the pre-training code. All scripts under there can be called as follows:
-
-```./scripts/launch_<name>.sh <path_to_config>```, where ```<path_to_config>``` points to a model configuration under ```configs```.
-
-## Pushing your model to HuggingFace Hub
-
-One question asks you to push your model to the huggingface hub. Steps:
-
-1- Create an account. In your account, create a **public** repo for your model. It's handle will be your_username/model_name.  
-
-2- cd to the model directory you want to push   
-
-3- run ```huggingface-cli upload your_username/model_name .```  
-
-## Submission Checklist and Instructions
-
-Run `bash create_submission.sh` then follow the instructions.
-
-Add and finalize all your files in the `submissions/` folder.
-
-Run `bash zip_submission.sh` and upload the relevant zip file to AutoGrader. Your submission folder should look like:
-
-```
-submission/
-├── configs/
-├── model.txt
-├── lm_inference.py
-└── report.pdf
-```