RichardErkhov/DeepMount00_-_Mistral-Ita-7b-gguf

Quantization made by Richard Erkhov.

Github

Discord

Request more models

Mistral-Ita-7b - GGUF

• Model creator: https://huggingface.co/DeepMount00/
• Original model: https://huggingface.co/DeepMount00/Mistral-Ita-7b/

Name	Quant method	Size
Mistral-Ita-7b.Q2_K.gguf	Q2_K	2.53GB
Mistral-Ita-7b.IQ3_XS.gguf	IQ3_XS	2.81GB
Mistral-Ita-7b.IQ3_S.gguf	IQ3_S	2.96GB
Mistral-Ita-7b.Q3_K_S.gguf	Q3_K_S	2.95GB
Mistral-Ita-7b.IQ3_M.gguf	IQ3_M	3.06GB
Mistral-Ita-7b.Q3_K.gguf	Q3_K	3.28GB
Mistral-Ita-7b.Q3_K_M.gguf	Q3_K_M	3.28GB
Mistral-Ita-7b.Q3_K_L.gguf	Q3_K_L	3.56GB
Mistral-Ita-7b.IQ4_XS.gguf	IQ4_XS	3.67GB
Mistral-Ita-7b.Q4_0.gguf	Q4_0	3.83GB
Mistral-Ita-7b.IQ4_NL.gguf	IQ4_NL	3.87GB
Mistral-Ita-7b.Q4_K_S.gguf	Q4_K_S	3.86GB
Mistral-Ita-7b.Q4_K.gguf	Q4_K	4.07GB
Mistral-Ita-7b.Q4_K_M.gguf	Q4_K_M	4.07GB
Mistral-Ita-7b.Q4_1.gguf	Q4_1	4.24GB
Mistral-Ita-7b.Q5_0.gguf	Q5_0	4.65GB
Mistral-Ita-7b.Q5_K_S.gguf	Q5_K_S	4.65GB
Mistral-Ita-7b.Q5_K.gguf	Q5_K	4.78GB
Mistral-Ita-7b.Q5_K_M.gguf	Q5_K_M	4.78GB
Mistral-Ita-7b.Q5_1.gguf	Q5_1	5.07GB
Mistral-Ita-7b.Q6_K.gguf	Q6_K	5.53GB

Original model description:

language: - it license: apache-2.0 tags: - text-generation-inference - text generation datasets: - DeepMount00/llm_ita_ultra

Mistral-7B-v0.1 for Italian Language Text Generation

Model Architecture

• Base Model: Mistral-7B-v0.1
• Specialization: Italian Language

Evaluation

For a detailed comparison of model performance, check out the Leaderboard for Italian Language Models.

Here's a breakdown of the performance metrics:

Metric	hellaswag_it acc_norm	arc_it acc_norm	m_mmlu_it 5-shot acc	Average
Accuracy Normalized	0.6731	0.5502	0.5364	0.5866

---

Quantized 4-Bit Version Available

A quantized 4-bit version of the model is available for use. This version offers a more efficient processing capability by reducing the precision of the model's computations to 4 bits, which can lead to faster performance and decreased memory usage. This might be particularly useful for deploying the model on devices with limited computational power or memory resources.

For more details and to access the model, visit the following link: Mistral-Ita-7b-GGUF 4-bit version.

---

How to Use

How to utilize my Mistral for Italian text generation

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

MODEL_NAME = "DeepMount00/Mistral-Ita-7b"

model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype=torch.bfloat16).eval()
model.to(device)
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)

def generate_answer(prompt):
    messages = [
        {"role": "user", "content": prompt},
    ]
    model_inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(device)
    generated_ids = model.generate(model_inputs, max_new_tokens=200, do_sample=True,
                                          temperature=0.001, eos_token_id=tokenizer.eos_token_id)
    decoded = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
    return decoded[0]

prompt = "Come si apre un file json in python?"
answer = generate_answer(prompt)
print(answer)

---

Developer

[Michele Montebovi]