Update README.md

README.md

@@ -16,40 +16,17 @@ base_model:
 ---
 # Model Card for Model ID
 
-<!-- Provide a quick summary of what the model is/does. -->
+# 🚀 Introducing Llama-3.2-1B-Instruct-Open-R1-Distill  
 
-Introducing Llama-3.2-1B-Instruct-Open-R1-Distill based on Llama-3.2-1B-Instruct and HuggingFace's [OpenR1](https://github.com/huggingface/open-r1) which is the fully open reproduction of DeepSeek-R1.
+Built on **Llama-3.2-1B-Instruct** and Hugging Face’s [OpenR1](https://github.com/huggingface/open-r1) — a fully open reproduction of **DeepSeek-R1** — this model brings powerful reasoning capabilities to compact, efficient architectures.  
 
-I have always been passionate about advancing state-of-the-art LLM technology in smaller models that can run efficiently on laptop CPUs and smartphones.
-Thanks to the recent breakthrough of DeepSeek-R1, it has become surprisingly easy to develop a reasoning model through distillation,
-requiring only supervised fine-tuning (SFT) on a dataset generated by a teacher model.
-Additionally, with the tools provided by Hugging Face, we now have a streamlined method to achieve this.
+## 📌 Why This Matters  
 
-To reprdouce the results, simply go to HuggingFace's [OpenR1](https://github.com/huggingface/open-r1) and install the package.
-
-And then execute the following command:
-```
-ACCELERATE_LOG_LEVEL=info accelerate launch --config_file recipes/accelerate_configs/zero3.yaml src/open_r1/sft.py --config recipes/config_llama3_instrcut_1b.yaml
-```
-
-You can create your own ```recipes/config_llama3_instrcut_1b.yaml``` by copying (config_full.yaml)[https://github.com/huggingface/open-r1/blob/main/recipes/qwen/Qwen2.5-1.5B-Instruct/sft/config_full.yaml]
-to the desired folder and change model path to ```model_name_or_path: meta-llama/Llama-3.2-1B-Instruct``` or any HuggingFace model repo id you are interested in. 
-You may also choose to training for more than 1 epoch (I trained for 5 epoch).
-Also, if you want to get intermediate checkpoints, set the save parameters accordingly:
-
-```
-save_strategy: "steps"
-save_steps: 100
-```
-
-I have tried to use 1 for both train and eval batch size on 1 Nvidia 4090 but still got OOM so I rented 4 x LS40s from [vast.ai]. Training 5 epoch only required < 4 hours.
-```
-per_device_eval_batch_size: 4
-per_device_train_batch_size: 4
-```
+I have always been passionate about pushing the boundaries of **LLM** technology in smaller models that can run seamlessly on laptop CPUs and smartphones.  
 
+With the recent breakthrough of **DeepSeek-R1**, developing a high-quality reasoning model through distillation has become remarkably straightforward. It requires only **supervised fine-tuning (SFT)** on a dataset generated by a teacher model.  
 
-## Model Details
+Thanks to **Hugging Face**, we now have a streamlined framework to make this process more accessible than ever.  
 
 ### Model Description
 
@@ -63,9 +40,16 @@ per_device_train_batch_size: 4
 - **License:** Apache License 2.0
 - **Finetuned from model [optional]:** Llama-3.2-1B-Instruct
 
-## Uses
+## 🎯 Uses
 
-```
+- 💡 **On-device AI assistants** for reasoning and general-purpose tasks  
+- 📱 **Mobile and edge AI applications** requiring lightweight models  
+- 🤖 **Chatbots and virtual assistants** optimized for efficiency  
+- 🏗 **Fine-tuning for specific domains** with SFT training
+
+### How to run the code?
+
+```{python}
 model = LlamaForCausalLM.from_pretrained("keeeeenw/Llama-3.2-1B-Instruct-Open-R1-Distill")
 
 # Prompt supported by HuggingFaceH4/Bespoke-Stratos-17k
@@ -92,163 +76,48 @@ outputs = model.generate(inputs['input_ids'],
 print(tokenizer.decode(outputs[0]))
 ```
 
-### Direct Use
-
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
-
-[More Information Needed]
-
-### Downstream Use [optional]
-
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
-
-[More Information Needed]
-
-### Out-of-Scope Use
-
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
-
-[More Information Needed]
-
-## Bias, Risks, and Limitations
-
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-
-[More Information Needed]
-
-### Recommendations
-
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-
-## How to Get Started with the Model
-
-Use the code below to get started with the model.
-
-[More Information Needed]
-
-## Training Details
-
-### Training Data
-
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-
-[More Information Needed]
-
-### Training Procedure
-
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-
-#### Preprocessing [optional]
-
-[More Information Needed]
-
-
-#### Training Hyperparameters
-
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-
-#### Speeds, Sizes, Times [optional]
-
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-
-[More Information Needed]
-
-## Evaluation
+## 🏋️‍♂️ Training Details
 
-<!-- This section describes the evaluation protocols and provides the results. -->
-
-### Testing Data, Factors & Metrics
-
-#### Testing Data
-
-<!-- This should link to a Dataset Card if possible. -->
-
-[More Information Needed]
-
-#### Factors
-
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-
-[More Information Needed]
-
-#### Metrics
-
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-
-[More Information Needed]
-
-### Results
-
-[More Information Needed]
-
-#### Summary
-
-
-
-## Model Examination [optional]
-
-<!-- Relevant interpretability work for the model goes here -->
-
-[More Information Needed]
-
-## Environmental Impact
-
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-
-## Technical Specifications [optional]
-
-### Model Architecture and Objective
-
-[More Information Needed]
-
-### Compute Infrastructure
-
-[More Information Needed]
-
-#### Hardware
-
-[More Information Needed]
-
-#### Software
-
-[More Information Needed]
-
-## Citation [optional]
-
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-
-**BibTeX:**
-
-[More Information Needed]
-
-**APA:**
-
-[More Information Needed]
+To reprdouce the results, simply go to HuggingFace's [OpenR1](https://github.com/huggingface/open-r1) and install the package.
 
-## Glossary [optional]
+And then execute the following command:
+```
+ACCELERATE_LOG_LEVEL=info accelerate launch --config_file recipes/accelerate_configs/zero3.yaml src/open_r1/sft.py --config recipes/config_llama3_instrcut_1b.yaml
+```
 
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+You can create your own ```recipes/config_llama3_instrcut_1b.yaml``` by copying [config_full.yaml](https://github.com/huggingface/open-r1/blob/main/recipes/qwen/Qwen2.5-1.5B-Instruct/sft/config_full.yaml)
+to the desired folder and change model path to ```model_name_or_path: meta-llama/Llama-3.2-1B-Instruct``` or any HuggingFace model repo id you are interested in. 
+You may also choose to training for more than 1 epoch (I trained for 5 epoch).
+Also, if you want to get intermediate checkpoints, set the save parameters accordingly:
 
-[More Information Needed]
+```
+save_strategy: "steps"
+save_steps: 100
+```
 
-## More Information [optional]
+I have tried to use 1 for both train and eval batch size on 1 Nvidia 4090 but still got OOM so I rented 4 x LS40s from [vast.ai]. Training 5 epoch only required < 4 hours.
+```
+per_device_eval_batch_size: 4
+per_device_train_batch_size: 4
+```
 
-[More Information Needed]
+## 📊 Evaluation
 
-## Model Card Authors [optional]
+The evaluation of this model is based on HuggingFace's instructions [OpenR1](https://github.com/huggingface/open-r1) 
 
-[More Information Needed]
+```
+NUM_GPUS=4
+MODEL="/root/open-r1/data/meta-llama/Llama-3.2-1B-Instruct"
+MODEL_ARGS="pretrained=$MODEL,dtype=float16,data_parallel_size=$NUM_GPUS,max_model_length=32768,gpu_memory_utilisation=0.8"
+TASK=aime24
+OUTPUT_DIR=data/evals/$MODEL
+
+lighteval vllm $MODEL_ARGS "custom|$TASK|0|0" \
+    --custom-tasks src/open_r1/evaluate.py \
+    --use-chat-template \
+    --system-prompt="Please reason step by step, and put your final answer within \boxed{}." \
+    --output-dir $OUTPUT_DIR
+```
 
-## Model Card Contact
+Results: To be added
 
-[More Information Needed]