Hugging Face's logo

MaziyarPanahi
/
calme-2.4-llama3-70b

Text Generation
Transformers
Safetensors
PyTorch
English
llama
axolotl
finetune
dpo
facebook
meta
llama-3
chatml
conversational
Eval Results
text-generation-inference
Model card Files Community
15
calme-2.4-llama3-70b / README.md
MaziyarPanahi's picture
MaziyarPanahi
Update README.md
c3c1a83 verified
preview code
|
raw
history blame
12.4 kB
metadata
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
Llama-3 DPO Logo

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

🏆 Open LLM Leaderboard Evaluation Results

Detailed results can be found here

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
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 Llama-3-70B finet-tuned models

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.

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.