File size: 12,647 Bytes

---
language:
- en
license: llama3
library_name: transformers
tags:
- axolotl
- finetune
- dpo
- facebook
- meta
- pytorch
- llama
- llama-3
- chatml
base_model: meta-llama/Meta-Llama-3-70B-Instruct
datasets:
- argilla/ultrafeedback-binarized-preferences
pipeline_tag: text-generation
license_name: llama3
license_link: LICENSE
inference: false
model_creator: MaziyarPanahi
quantized_by: MaziyarPanahi
model-index:
- name: Llama-3-70B-Instruct-DPO-v0.4
  results:
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: AI2 Reasoning Challenge (25-Shot)
      type: ai2_arc
      config: ARC-Challenge
      split: test
      args:
        num_few_shot: 25
    metrics:
    - type: acc_norm
      value: 72.61
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: HellaSwag (10-Shot)
      type: hellaswag
      split: validation
      args:
        num_few_shot: 10
    metrics:
    - type: acc_norm
      value: 86.03
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MMLU (5-Shot)
      type: cais/mmlu
      config: all
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 80.5
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: TruthfulQA (0-shot)
      type: truthful_qa
      config: multiple_choice
      split: validation
      args:
        num_few_shot: 0
    metrics:
    - type: mc2
      value: 63.26
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: Winogrande (5-shot)
      type: winogrande
      config: winogrande_xl
      split: validation
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 83.58
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: GSM8k (5-shot)
      type: gsm8k
      config: main
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 87.34
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: IFEval (0-Shot)
      type: HuggingFaceH4/ifeval
      args:
        num_few_shot: 0
    metrics:
    - type: inst_level_strict_acc and prompt_level_strict_acc
      value: 50.27
      name: strict accuracy
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: BBH (3-Shot)
      type: BBH
      args:
        num_few_shot: 3
    metrics:
    - type: acc_norm
      value: 48.4
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MATH Lvl 5 (4-Shot)
      type: hendrycks/competition_math
      args:
        num_few_shot: 4
    metrics:
    - type: exact_match
      value: 22.66
      name: exact match
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: GPQA (0-shot)
      type: Idavidrein/gpqa
      args:
        num_few_shot: 0
    metrics:
    - type: acc_norm
      value: 11.97
      name: acc_norm
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MuSR (0-shot)
      type: TAUR-Lab/MuSR
      args:
        num_few_shot: 0
    metrics:
    - type: acc_norm
      value: 13.1
      name: acc_norm
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MMLU-PRO (5-shot)
      type: TIGER-Lab/MMLU-Pro
      config: main
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 46.71
      name: accuracy
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4
      name: Open LLM Leaderboard
---

<img src="./llama-3-merges.webp" alt="Llama-3 DPO Logo" width="500" style="margin-left:'auto' margin-right:'auto' display:'block'"/>


# MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4

This model is a fine-tune (DPO) of `meta-llama/Meta-Llama-3-70B-Instruct` model.

# ⚡ Quantized GGUF

All GGUF models are available here: [MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4-GGUF](https://huggingface.co/MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4-GGUF)

# 🏆 [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_MaziyarPanahi__Llama-3-70B-Instruct-DPO-v0.4)

|             Metric              |Value|
|---------------------------------|----:|
|Avg.                             |78.89|
|AI2 Reasoning Challenge (25-Shot)|72.61|
|HellaSwag (10-Shot)              |86.03|
|MMLU (5-Shot)                    |80.50|
|TruthfulQA (0-shot)              |63.26|
|Winogrande (5-shot)              |83.58|
|GSM8k (5-shot)                   |87.34|

**Top 10 models on the Leaderboard**
<img src="./llama-3-70b-top-10.png" alt="Llama-3-70B finet-tuned models" style="margin-left:'auto' margin-right:'auto' display:'block'"/>


# Prompt Template

This model uses `ChatML` prompt template:

```
<|im_start|>system
{System}
<|im_end|>
<|im_start|>user
{User}
<|im_end|>
<|im_start|>assistant
{Assistant}
````

# How to use

You can use this model by using `MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4` as the model name in Hugging Face's
transformers library.

```python
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
from transformers import pipeline
import torch

model_id = "MaziyarPanahi/Llama-3-70B-Instruct-DPO-v0.4"

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True,
    # attn_implementation="flash_attention_2"
)

tokenizer = AutoTokenizer.from_pretrained(
    model_id,
    trust_remote_code=True
)

streamer = TextStreamer(tokenizer)

pipeline = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    model_kwargs={"torch_dtype": torch.bfloat16},
    streamer=streamer
)

# Then you can use the pipeline to generate text.

messages = [
    {"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"},
    {"role": "user", "content": "Who are you?"},
]

prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

terminators = [
    tokenizer.eos_token_id,
    tokenizer.convert_tokens_to_ids("<|im_end|>"),
    tokenizer.convert_tokens_to_ids("<|eot_id|>") # safer to have this too
]

outputs = pipeline(
    prompt,
    max_new_tokens=2048,
    eos_token_id=terminators,
    do_sample=True,
    temperature=0.6,
    top_p=0.95,
)
print(outputs[0]["generated_text"][len(prompt):])
```

## Sample Response

```
<|im_start|>system
You are a helpful assistant.<|im_end|>
<|im_start|>user
describe about pros and cons of docker system.<|im_end|>
<|im_start|>assistant
```

```
Here are the pros and cons of the Docker system:

**Pros:**

1. **Lightweight**: Docker containers are lightweight and use fewer resources compared to traditional virtual machines.
2. **Isolation**: Docker provides a high level of isolation between containers, ensuring that each application runs independently without interfering with others.
3. **Portability**: Docker containers are highly portable and can run on any system that supports Docker, without requiring specific dependencies or configurations.
4. **Efficient Use of Resources**: Docker containers can share the same kernel and use fewer resources, making them more efficient than traditional virtual machines.
5. **Fast Deployment**: Docker containers can be spun up quickly, making it ideal for rapid deployment and scaling of applications.
6. **Easy Maintenance**: Docker provides a simple way to manage and update applications, with features like rolling updates and rollbacks.
7. **Security**: Docker provides a secure environment for applications, with features like network isolation and encryption.
8. **Large Ecosystem**: Docker has a large and active ecosystem, with a wide range of tools and plugins available for various tasks.
9. **Version Control**: Docker provides an easy way to manage different versions of an application, with features like version tagging and rollbacks.
10. **Cost-Effective**: Docker can help reduce costs by minimizing the need for virtual machines and reducing the overhead of resource allocation.

**Cons:**

1. **Steep Learning Curve**: Docker has a complex architecture and requires a good understanding of containerization, networking, and storage.
2. **Overhead**: While Docker containers are lightweight, they still require some overhead in terms of resources and management.
3. **Security Risks**: If not properly configured, Docker containers can introduce security risks, such as unauthorized access to the host system.
4. **Limited Support for GUI Applications**: Docker is primarily designed for running command-line applications, and running GUI applications can be challenging.
5. **Networking Complexity**: Docker's networking model can be complex, especially in large-scale deployments.
6. **Storage Complexity**: Docker's storage model can be complex, especially when dealing with persistent data.
7. **Limited Support for Legacy Applications**: Docker is designed for modern, cloud-native applications, and may not be suitable for legacy applications that rely on older technologies.
8. **Dependence on Docker Hub**: Docker relies heavily on Docker Hub, which can be a single point of failure if the hub is down or experiencing issues.
9. **Orchestration Complexity**: Docker provides a built-in orchestration tool, Docker Swarm, but it can be complex to set up and manage.
10. **Vendor Lock-in**: Docker is a proprietary technology, and while it has a large ecosystem, it can be difficult to switch to alternative containerization platforms.

Overall, Docker provides a powerful and flexible way to deploy and manage applications, but it requires careful planning, configuration, and management to ensure optimal performance and security.
```

# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_MaziyarPanahi__Llama-3-70B-Instruct-DPO-v0.4)

|      Metric       |Value|
|-------------------|----:|
|Avg.               |32.18|
|IFEval (0-Shot)    |50.27|
|BBH (3-Shot)       |48.40|
|MATH Lvl 5 (4-Shot)|22.66|
|GPQA (0-shot)      |11.97|
|MuSR (0-shot)      |13.10|
|MMLU-PRO (5-shot)  |46.71|