|
--- |
|
license: apache-2.0 |
|
tags: |
|
- merge |
|
- mergekit |
|
- lazymergekit |
|
- bfloat16 |
|
- text-generation-inference |
|
- model_stock |
|
- crypto |
|
- finance |
|
- llama |
|
language: |
|
- en |
|
base_model: |
|
- Chainbase-Labs/Theia-Llama-3.1-8B-v1 |
|
- EpistemeAI/Fireball-Meta-Llama-3.2-8B-Instruct-agent-003-128k-code-DPO |
|
- mukaj/Llama-3.1-Hawkish-8B |
|
pipeline_tag: text-generation |
|
library_name: transformers |
|
--- |
|
|
|
# ZeroXClem/LLama3.1-Hawkish-Theia-Fireball-8B |
|
|
|
**ZeroXClem/LLama3.1-Hawkish-Theia-Fireball-8B** is an advanced language model meticulously crafted by merging three pre-trained models using the powerful [mergekit](https://github.com/cg123/mergekit) framework. This fusion leverages the **Model Stock** merge method to combine the specialized capabilities of **Theia-Llama**, **Fireball-Meta-Llama**, and **Llama-Hawkish**. The resulting model excels in creative text generation, technical instruction following, financial reasoning, and dynamic conversational interactions. |
|
|
|
## π Merged Models |
|
|
|
This model merge incorporates the following: |
|
|
|
- [**Chainbase-Labs/Theia-Llama-3.1-8B-v1**](https://huggingface.co/Chainbase-Labs/Theia-Llama-3.1-8B-v1): Specializes in cryptocurrency-oriented knowledge, enhancing the model's ability to generate and comprehend crypto-related content with high accuracy and depth. |
|
|
|
- [**EpistemeAI/Fireball-Meta-Llama-3.2-8B-Instruct-agent-003-128k-code-DPO**](https://huggingface.co/EpistemeAI/Fireball-Meta-Llama-3.2-8B-Instruct-agent-003-128k-code-DPO): Focuses on instruction-following and coding capabilities, improving the model's performance in understanding and executing user commands, as well as generating executable code snippets. |
|
|
|
- [**mukaj/Llama-3.1-Hawkish-8B**](https://huggingface.co/mukaj/Llama-3.1-Hawkish-8B): Enhances financial reasoning and mathematical precision, enabling the model to handle complex financial analyses, economic discussions, and quantitative problem-solving with high proficiency. |
|
|
|
## 𧩠Merge Configuration |
|
|
|
The configuration below outlines how the models are merged using the **Model Stock** method. This approach ensures a balanced and effective integration of the unique strengths from each source model. |
|
|
|
```yaml |
|
# Merge configuration for ZeroXClem/LLama3.1-Hawkish-Theia-Fireball-8B using Model Stock |
|
|
|
models: |
|
- model: Chainbase-Labs/Theia-Llama-3.1-8B-v1 |
|
- model: EpistemeAI/Fireball-Meta-Llama-3.2-8B-Instruct-agent-003-128k-code-DPO |
|
- model: mukaj/Llama-3.1-Hawkish-8B |
|
merge_method: model_stock |
|
base_model: mukaj/Llama-3.1-Hawkish-8B |
|
normalize: false |
|
int8_mask: true |
|
dtype: bfloat16 |
|
``` |
|
|
|
### Key Parameters |
|
|
|
- **Merge Method (`merge_method`):** Utilizes the **Model Stock** method, as described in [Model Stock](https://arxiv.org/abs/2403.19522), to effectively combine multiple models by leveraging their strengths. |
|
|
|
- **Models (`models`):** Specifies the list of models to be merged: |
|
- **Chainbase-Labs/Theia-Llama-3.1-8B-v1:** Enhances cryptocurrency-oriented knowledge and content generation. |
|
- **EpistemeAI/Fireball-Meta-Llama-3.2-8B-Instruct-agent-003-128k-code-DPO:** Improves instruction-following and coding capabilities. |
|
- **mukaj/Llama-3.1-Hawkish-8B:** Enhances financial reasoning and mathematical precision. |
|
|
|
- **Base Model (`base_model`):** Defines the foundational model for the merge, which is **mukaj/Llama-3.1-Hawkish-8B** in this case. |
|
|
|
- **Normalization (`normalize`):** Set to `false` to retain the original scaling of the model weights during the merge. |
|
|
|
- **INT8 Mask (`int8_mask`):** Enabled (`true`) to apply INT8 quantization masking, optimizing the model for efficient inference without significant loss in precision. |
|
|
|
- **Data Type (`dtype`):** Uses `bfloat16` to maintain computational efficiency while ensuring high precision. |
|
|
|
## π Performance Highlights |
|
|
|
- **Cryptocurrency Knowledge:** Enhanced ability to generate and comprehend crypto-related content, making the model highly effective for blockchain discussions, crypto market analysis, and related queries. |
|
|
|
- **Instruction Following and Coding:** Improved performance in understanding and executing user instructions, as well as generating accurate and executable code snippets, suitable for coding assistance and technical support. |
|
|
|
- **Financial Reasoning and Mathematical Precision:** Advanced capabilities in handling complex financial analyses, economic discussions, and quantitative problem-solving, making the model ideal for financial modeling, investment analysis, and educational purposes. |
|
|
|
- **Smooth Weight Blending:** Utilization of the Model Stock method ensures a harmonious integration of different model attributes, resulting in balanced performance across various specialized tasks. |
|
|
|
- **Optimized Inference:** INT8 masking and `bfloat16` data type contribute to efficient computation, enabling faster response times without compromising quality. |
|
|
|
## π― Use Case & Applications |
|
|
|
**ZeroXClem/LLama3.1-Hawkish-Theia-Fireball-8B** is designed to excel in environments that demand a combination of creative generation, technical instruction following, financial reasoning, and dynamic conversational interactions. Ideal applications include: |
|
|
|
- **Cryptocurrency Analysis and Reporting:** Generating detailed reports, analyses, and summaries related to blockchain projects, crypto markets, and financial technologies. |
|
|
|
- **Coding Assistance and Technical Support:** Providing accurate and executable code snippets, debugging assistance, and technical explanations for developers and technical professionals. |
|
|
|
- **Financial Modeling and Investment Analysis:** Assisting financial analysts and investors in creating models, performing economic analyses, and making informed investment decisions through precise calculations and reasoning. |
|
|
|
- **Educational Tools and Tutoring Systems:** Offering detailed explanations, answering complex questions, and assisting in educational content creation across subjects like finance, economics, and mathematics. |
|
|
|
- **Interactive Conversational Agents:** Powering chatbots and virtual assistants with specialized knowledge in cryptocurrency, finance, and technical domains, enhancing user interactions and support. |
|
|
|
- **Content Generation for Finance and Tech Blogs:** Creating high-quality, contextually relevant content for blogs, articles, and marketing materials focused on finance, technology, and cryptocurrency. |
|
|
|
## π Usage |
|
|
|
To utilize **ZeroXClem/LLama3.1-Hawkish-Theia-Fireball-8B**, follow the steps below: |
|
|
|
### Installation |
|
|
|
First, install the necessary libraries: |
|
|
|
```bash |
|
pip install -qU transformers accelerate |
|
``` |
|
|
|
### Example Code |
|
|
|
Below is an example of how to load and use the model for text generation: |
|
|
|
```python |
|
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline |
|
import torch |
|
|
|
# Define the model name |
|
model_name = "ZeroXClem/LLama3.1-Hawkish-Theia-Fireball-8B" |
|
|
|
# Load the tokenizer |
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
|
|
# Load the model |
|
model = AutoModelForCausalLM.from_pretrained( |
|
model_name, |
|
torch_dtype=torch.bfloat16, |
|
device_map="auto" |
|
) |
|
|
|
# Initialize the pipeline |
|
text_generator = pipeline( |
|
"text-generation", |
|
model=model, |
|
tokenizer=tokenizer, |
|
torch_dtype=torch.bfloat16, |
|
device_map="auto" |
|
) |
|
|
|
# Define the input prompt |
|
prompt = "Explain the impact of decentralized finance on traditional banking systems." |
|
|
|
# Generate the output |
|
outputs = text_generator( |
|
prompt, |
|
max_new_tokens=150, |
|
do_sample=True, |
|
temperature=0.7, |
|
top_k=50, |
|
top_p=0.95 |
|
) |
|
|
|
# Print the generated text |
|
print(outputs[0]["generated_text"]) |
|
``` |
|
|
|
### Notes |
|
|
|
- **Fine-Tuning:** This merged model may require fine-tuning to optimize performance for specific applications or domains, especially in highly specialized fields like cryptocurrency and finance. |
|
|
|
- **Resource Requirements:** Ensure that your environment has sufficient computational resources, especially GPU-enabled hardware, to handle the model efficiently during inference. |
|
|
|
- **Customization:** Users can adjust parameters such as `temperature`, `top_k`, and `top_p` to control the creativity and diversity of the generated text, tailoring the model's output to specific needs. |
|
|
|
|
|
## π License |
|
|
|
This model is open-sourced under the **Apache-2.0 License**. |
|
|
|
## π‘ Tags |
|
|
|
- `merge` |
|
- `mergekit` |
|
- `model_stock` |
|
- `Llama` |
|
- `Hawkish` |
|
- `Theia` |
|
- `Fireball` |
|
- `ZeroXClem/LLama3.1-Hawkish-Theia-Fireball-8B` |
|
- `Chainbase-Labs/Theia-Llama-3.1-8B-v1` |
|
- `EpistemeAI/Fireball-Meta-Llama-3.2-8B-Instruct-agent-003-128k-code-DPO` |
|
- `mukaj/Llama-3.1-Hawkish-8B` |