🚀 al-baka-llama3-8b
Al Baka is an Experimental Fine Tuned Model based on the new released LLAMA3-8B Model on the Stanford Alpaca dataset Arabic version Yasbok/Alpaca_arabic_instruct.
Model Summary
- Model Type: Llama3-8B FineTuned Model
- Language(s): Arabic
- Base Model: LLAMA-3-8B
- Dataset: Yasbok/Alpaca_arabic_instruct
Model Details
-
The model was fine-tuned in 4-bit precision using unsloth
-
The run is performed only for 1000 steps with a single Google Colab T4 GPU NVIDIA GPU with 15 GB of available memory.
The model is currently being Experimentally Fine Tuned to assess LLaMA-3's response to Arabic, following a brief period of fine-tuning. Larger and more sophisticated models will be introduced soon.
How to Get Started with the Model
Setup
# Install packages
%%capture
import torch
major_version, minor_version = torch.cuda.get_device_capability()
!pip install "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
if major_version >= 8:
# Use this for new GPUs like Ampere, Hopper GPUs (RTX 30xx, RTX 40xx, A100, H100, L40)
!pip install --no-deps packaging ninja einops flash-attn xformers trl peft accelerate bitsandbytes
else:
# Use this for older GPUs (V100, Tesla T4, RTX 20xx)
!pip install --no-deps xformers trl peft accelerate bitsandbytes
pass
First, Load the Model
from unsloth import FastLanguageModel
import torch
max_seq_length = 2048 # Choose any! We auto support RoPE Scaling internally!
dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.
model, tokenizer = FastLanguageModel.from_pretrained(
model_name = "Omartificial-Intelligence-Space/al-baka-16bit-llama3-8b",
max_seq_length = max_seq_length,
dtype = dtype,
load_in_4bit = load_in_4bit,
# token = "hf_...", # use one if using gated models like meta-llama/Llama-2-7b-hf
)
Second, Try the model
alpaca_prompt = """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
### Instruction:
{}
### Input:
{}
### Response:
{}"""
# alpaca_prompt = Copied from above
FastLanguageModel.for_inference(model) # Enable native 2x faster inference
inputs = tokenizer(
[
alpaca_prompt.format(
"استخدم البيانات المعطاة لحساب الوسيط.", # instruction
"[2 ، 3 ، 7 ، 8 ، 10]", # input
"", # output - leave this blank for generation!
)
], return_tensors = "pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens = 64, use_cache = True)
tokenizer.batch_decode(outputs)
Recommendations
- unsloth for finetuning models. You can get a 2x faster finetuned model which can be exported to any format or uploaded to Hugging Face.