Quantization made by Richard Erkhov.

cria-llama2-7b-v1.3 - GGUF

Model creator: https://huggingface.co/davzoku/
Original model: https://huggingface.co/davzoku/cria-llama2-7b-v1.3/

Name	Quant method	Size
cria-llama2-7b-v1.3.Q2_K.gguf	Q2_K	2.36GB
cria-llama2-7b-v1.3.IQ3_XS.gguf	IQ3_XS	2.6GB
cria-llama2-7b-v1.3.IQ3_S.gguf	IQ3_S	2.75GB
cria-llama2-7b-v1.3.Q3_K_S.gguf	Q3_K_S	2.75GB
cria-llama2-7b-v1.3.IQ3_M.gguf	IQ3_M	2.9GB
cria-llama2-7b-v1.3.Q3_K.gguf	Q3_K	3.07GB
cria-llama2-7b-v1.3.Q3_K_M.gguf	Q3_K_M	3.07GB
cria-llama2-7b-v1.3.Q3_K_L.gguf	Q3_K_L	3.35GB
cria-llama2-7b-v1.3.IQ4_XS.gguf	IQ4_XS	3.4GB
cria-llama2-7b-v1.3.Q4_0.gguf	Q4_0	3.56GB
cria-llama2-7b-v1.3.IQ4_NL.gguf	IQ4_NL	3.58GB
cria-llama2-7b-v1.3.Q4_K_S.gguf	Q4_K_S	3.59GB
cria-llama2-7b-v1.3.Q4_K.gguf	Q4_K	3.8GB
cria-llama2-7b-v1.3.Q4_K_M.gguf	Q4_K_M	3.8GB
cria-llama2-7b-v1.3.Q4_1.gguf	Q4_1	3.95GB
cria-llama2-7b-v1.3.Q5_0.gguf	Q5_0	4.33GB
cria-llama2-7b-v1.3.Q5_K_S.gguf	Q5_K_S	4.33GB
cria-llama2-7b-v1.3.Q5_K.gguf	Q5_K	4.45GB
cria-llama2-7b-v1.3.Q5_K_M.gguf	Q5_K_M	4.45GB
cria-llama2-7b-v1.3.Q5_1.gguf	Q5_1	4.72GB
cria-llama2-7b-v1.3.Q6_K.gguf	Q6_K	5.15GB
cria-llama2-7b-v1.3.Q8_0.gguf	Q8_0	6.67GB

Original model description:

inference: false language: en license: llama2 model_type: llama datasets: - mlabonne/CodeLlama-2-20k pipeline_tag: text-generation tags: - llama-2

CRIA v1.3

💡 Article | 💻 Github | 📔 Colab 1,2

What is CRIA?

krē-ə plural crias. : a baby llama, alpaca, vicuña, or guanaco.

or what ChatGPT suggests, "Crafting a Rapid prototype of an Intelligent llm App using open source resources".

The initial objective of the CRIA project is to develop a comprehensive end-to-end chatbot system, starting from the instruction-tuning of a large language model and extending to its deployment on the web using frameworks such as Next.js.

Specifically, we have fine-tuned the llama-2-7b-chat-hf model with QLoRA (4-bit precision) using the mlabonne/CodeLlama-2-20k dataset. This fine-tuned model serves as the backbone for the CRIA chat platform.

📦 Model Release

CRIA v1.3 comes with several variants.

davzoku/cria-llama2-7b-v1.3: Merged Model
davzoku/cria-llama2-7b-v1.3-GGML: Quantized Merged Model
davzoku/cria-llama2-7b-v1.3_peft: PEFT adapter

🔧 Training

It was trained on a Google Colab notebook with a T4 GPU and high RAM.

Training procedure

The following bitsandbytes quantization config was used during training:

load_in_8bit: False
load_in_4bit: True
llm_int8_threshold: 6.0
llm_int8_skip_modules: None
llm_int8_enable_fp32_cpu_offload: False
llm_int8_has_fp16_weight: False
bnb_4bit_quant_type: nf4
bnb_4bit_use_double_quant: False
bnb_4bit_compute_dtype: float16

Framework versions

PEFT 0.4.0

💻 Usage

# pip install transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "davzoku/cria-llama2-7b-v1.3"
prompt = "What is a cria?"

tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

sequences = pipeline(
    f'<s>[INST] {prompt} [/INST]',
    do_sample=True,
    top_k=10,
    num_return_sequences=1,
    eos_token_id=tokenizer.eos_token_id,
    max_length=200,
)
for seq in sequences:
    print(f"Result: {seq['generated_text']}")

References

We'd like to thank:

mlabonne for his article and resources on implementation of instruction tuning
TheBloke for his script for LLM quantization.