|
--- |
|
base_model: Spestly/Ava-1.0-12B |
|
library_name: transformers |
|
license: apache-2.0 |
|
datasets: |
|
- nvidia/HelpSteer2 |
|
tags: |
|
- unsloth |
|
- llama-cpp |
|
- gguf-my-repo |
|
--- |
|
|
|
# Triangle104/Ava-1.0-12B-Q4_K_S-GGUF |
|
This model was converted to GGUF format from [`Spestly/Ava-1.0-12B`](https://huggingface.co/Spestly/Ava-1.0-12B) using llama.cpp via the ggml.ai's [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space. |
|
Refer to the [original model card](https://huggingface.co/Spestly/Ava-1.0-12B) for more details on the model. |
|
|
|
--- |
|
Model details: |
|
- |
|
Ava 1.0 |
|
|
|
Ava 1.0 is a cutting-edge conversational AI model, fine-tuned from Mistral's NeMo to deliver exceptional conversational capabilities. Designed to be your go-to AI for engaging, accurate, and context-aware dialogues, Ava 1.0 incorporates updated knowledge and enhanced natural language understanding to provide an unparalleled user experience. |
|
|
|
Key Features |
|
- |
|
Enhanced Conversational Skills: Ava 1.0 demonstrates fluid and human-like dialogue generation with improved contextual understanding. |
|
Updated Knowledge Base: Trained on the latest datasets, Ava 1.0 ensures responses are relevant and informed. |
|
Multi-Turn Conversation: Handles complex, multi-turn interactions seamlessly, maintaining coherence and focus. |
|
Personalized Assistance: Adapts responses based on user preferences and context. |
|
Multilingual Support: Capable of understanding and responding in multiple languages with high accuracy. |
|
|
|
Why Ava 1.0? |
|
- |
|
Ava 1.0 is built to excel in a wide range of applications: |
|
|
|
Customer Support: Provides intelligent, empathetic, and accurate responses to customer queries. |
|
Education: Acts as an interactive tutor, offering explanations and personalized guidance. |
|
Personal Assistance: Supports daily tasks, scheduling, and answering general queries with ease. |
|
Creative Collaboration: Assists with brainstorming, writing, and other creative processes. |
|
|
|
Usage |
|
- |
|
Using Ava 1.0 in your project is straightforward. Here’s a quick setup guide: |
|
Installation |
|
|
|
Ensure you have the necessary libraries and dependencies installed. Use the following command: |
|
|
|
pip install transformers |
|
|
|
Implementation |
|
- |
|
Here’s a sample Python script to interact with Ava 1.0: |
|
|
|
# Use a pipeline as a high-level helper |
|
from transformers import pipeline |
|
|
|
pipe = pipeline("text-generation", model="Spestly/Ava-12B") |
|
|
|
#OR |
|
|
|
# Load model directly |
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
|
tokenizer = AutoTokenizer.from_pretrained("Spestly/Ava-12B") |
|
model = AutoModelForCausalLM.from_pretrained("Spestly/Ava-12B") |
|
|
|
Training Highlights |
|
- |
|
Ava 1.0 was fine-tuned with the following enhancements: |
|
|
|
Extensive Conversational Dataset: Leveraging a wide array of open-domain and specialized conversational datasets. |
|
Knowledge Integration: Incorporating recent advancements and updates to provide cutting-edge insights. |
|
Fine-Tuning on Mistral NeMo: Utilizing the powerful Mistral NeMo framework for robust and efficient training. |
|
|
|
Limitations |
|
- |
|
Contextual Challenges: In rare cases, Ava 1.0 may misinterpret ambiguous inputs. |
|
Hardware Requirements: Optimal performance requires a robust system with GPU acceleration. |
|
|
|
Roadmap |
|
- |
|
Ava 2.0: Introducing real-time learning capabilities and broader conversational adaptability. |
|
Lightweight Model: Developing a lightweight version optimized for edge devices. |
|
Domain-Specific Fine-Tunes: Specialized versions for industries like healthcare, education, and finance. |
|
|
|
License |
|
- |
|
Ava 1.0 is released under the Apache 2.0 license. |
|
|
|
Contact |
|
- |
|
For inquiries, feedback, or support, feel free to reach out: |
|
|
|
Email: aayan.mishra@proton.me |
|
GitHub: Spestly |
|
Website: Ava Project Page |
|
|
|
--- |
|
## Use with llama.cpp |
|
Install llama.cpp through brew (works on Mac and Linux) |
|
|
|
```bash |
|
brew install llama.cpp |
|
|
|
``` |
|
Invoke the llama.cpp server or the CLI. |
|
|
|
### CLI: |
|
```bash |
|
llama-cli --hf-repo Triangle104/Ava-1.0-12B-Q4_K_S-GGUF --hf-file ava-1.0-12b-q4_k_s.gguf -p "The meaning to life and the universe is" |
|
``` |
|
|
|
### Server: |
|
```bash |
|
llama-server --hf-repo Triangle104/Ava-1.0-12B-Q4_K_S-GGUF --hf-file ava-1.0-12b-q4_k_s.gguf -c 2048 |
|
``` |
|
|
|
Note: You can also use this checkpoint directly through the [usage steps](https://github.com/ggerganov/llama.cpp?tab=readme-ov-file#usage) listed in the Llama.cpp repo as well. |
|
|
|
Step 1: Clone llama.cpp from GitHub. |
|
``` |
|
git clone https://github.com/ggerganov/llama.cpp |
|
``` |
|
|
|
Step 2: Move into the llama.cpp folder and build it with `LLAMA_CURL=1` flag along with other hardware-specific flags (for ex: LLAMA_CUDA=1 for Nvidia GPUs on Linux). |
|
``` |
|
cd llama.cpp && LLAMA_CURL=1 make |
|
``` |
|
|
|
Step 3: Run inference through the main binary. |
|
``` |
|
./llama-cli --hf-repo Triangle104/Ava-1.0-12B-Q4_K_S-GGUF --hf-file ava-1.0-12b-q4_k_s.gguf -p "The meaning to life and the universe is" |
|
``` |
|
or |
|
``` |
|
./llama-server --hf-repo Triangle104/Ava-1.0-12B-Q4_K_S-GGUF --hf-file ava-1.0-12b-q4_k_s.gguf -c 2048 |
|
``` |
|
|
