muranAI/gemma-3n-E4B-it-GGUF
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

初始化项目，由ModelHub XC社区提供模型

Browse Source 
Model: muranAI/gemma-3n-E4B-it-GGUF
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-08-11 14:43:17 +08:00
commit f96ea87fe4
19 changed files with 954 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

50
.gitattributes vendored Normal file
View File

@@ -0,0 +1,50 @@
*.7z filter=lfs diff=lfs merge=lfs -text
*.arrow filter=lfs diff=lfs merge=lfs -text
*.bin filter=lfs diff=lfs merge=lfs -text
*.bz2 filter=lfs diff=lfs merge=lfs -text
*.ckpt filter=lfs diff=lfs merge=lfs -text
*.ftz filter=lfs diff=lfs merge=lfs -text
*.gz filter=lfs diff=lfs merge=lfs -text
*.h5 filter=lfs diff=lfs merge=lfs -text
*.joblib filter=lfs diff=lfs merge=lfs -text
*.lfs.* filter=lfs diff=lfs merge=lfs -text
*.mlmodel filter=lfs diff=lfs merge=lfs -text
*.model filter=lfs diff=lfs merge=lfs -text
*.msgpack filter=lfs diff=lfs merge=lfs -text
*.npy filter=lfs diff=lfs merge=lfs -text
*.npz filter=lfs diff=lfs merge=lfs -text
*.onnx filter=lfs diff=lfs merge=lfs -text
*.ot filter=lfs diff=lfs merge=lfs -text
*.parquet filter=lfs diff=lfs merge=lfs -text
*.pb filter=lfs diff=lfs merge=lfs -text
*.pickle filter=lfs diff=lfs merge=lfs -text
*.pkl filter=lfs diff=lfs merge=lfs -text
*.pt filter=lfs diff=lfs merge=lfs -text
*.pth filter=lfs diff=lfs merge=lfs -text
*.rar filter=lfs diff=lfs merge=lfs -text
*.safetensors filter=lfs diff=lfs merge=lfs -text
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
*.tar.* filter=lfs diff=lfs merge=lfs -text
*.tar filter=lfs diff=lfs merge=lfs -text
*.tflite filter=lfs diff=lfs merge=lfs -text
*.tgz filter=lfs diff=lfs merge=lfs -text
*.wasm filter=lfs diff=lfs merge=lfs -text
*.xz filter=lfs diff=lfs merge=lfs -text
*.zip filter=lfs diff=lfs merge=lfs -text
*.zst filter=lfs diff=lfs merge=lfs -text
*tfevents* filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-f16.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q2_k.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q3_k_l.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q3_k_m.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q3_k_s.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q4_0.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q4_1.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q4_k_m.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q4_k_s.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q5_0.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q5_1.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q5_k_m.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q5_k_s.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q6_k.gguf filter=lfs diff=lfs merge=lfs -text
gemma-3n-e4b-it-q8_0.gguf filter=lfs diff=lfs merge=lfs -text

837
README.md Normal file
View File

@@ -0,0 +1,837 @@
---
license: gemma
base_model: google/gemma-3n-e4b-it
tags:
- gemma3n
- gguf
- quantized
- llama.cpp
- ollama
- inference
- conversational
- multilingual
language:
- en
- multilingual
pipeline_tag: text-generation
model_type: gemma3n
library_name: transformers
---
<p style="margin-bottom: 0;">
<em>See <a href="https://huggingface.co/muranAI">our collection</a> for all new Models.</em>
</p>
<div style="display: flex; gap: 5px; align-items: center; ">
<a href="https://muranai.com/">
<img src="https://muranai.com/images/logo_white.png" width="133">
</a>
</div>
# Gemma 3n E4B IT - Complete GGUF Collection
This repository contains a comprehensive collection of **Gemma 3n E4B Instruction-Tuned** models quantized to various GGUF formats for efficient inference on different hardware configurations.
## 📊 **Quantization Overview**
| **Format** | **Size (GB)** | **Compression** | **Quality** | **Use Case** |
|------------|---------------|-----------------|-------------|--------------|
| **F16** | 13.0 | 1.0x | 🌟🌟🌟🌟🌟 | Research, maximum quality |
| **Q8_0** | 6.8 | 1.9x | 🌟🌟🌟🌟🌟 | Production, near-original quality |
| **Q6_K** | 5.3 | 2.5x | 🌟🌟🌟🌟 | High-quality inference |
| **Q5_1** | 4.9 | 2.6x | 🌟🌟🌟🌟 | Balanced quality/speed |
| **Q5_K_M** | 4.6 | 2.8x | 🌟🌟🌟🌟 | **Recommended for most users** |
| **Q5_K_S** | 4.5 | 2.8x | 🌟🌟🌟🌟 | Slightly faster Q5 |
| **Q5_0** | 4.5 | 2.8x | 🌟🌟🌟🌟 | Good balance |
| **Q4_1** | 4.2 | 3.1x | 🌟🌟🌟 | Better 4-bit quality |
| **Q4_K_M** | 3.9 | 3.3x | 🌟🌟🌟 | **Popular choice** |
| **Q4_K_S** | 3.8 | 3.4x | 🌟🌟🌟 | Efficient 4-bit |
| **Q4_0** | 3.8 | 3.4x | 🌟🌟🌟 | **Most compatible** |
| **Q3_K_L** | 3.4 | 3.8x | 🌟🌟 | High compression |
| **Q3_K_M** | 3.2 | 4.0x | 🌟🌟 | Aggressive compression |
| **Q3_K_S** | 3.0 | 4.3x | 🌟🌟 | Maximum speed |
| **Q2_K** | 2.6 | 5.0x | 🌟 | Extreme compression |
## 🚀 **Quick Start**
### Using with Ollama
```bash
# Option 1: Use pre-built model (requires Ollama ≥ 0.10.0)
ollama run gemma3n:e4b
# Option 2: Import custom quantization
ollama create my-gemma3n -f Modelfile
```
### Using with llama.cpp
```bash
# Download your preferred quantization
./llama-server -m gemma-3n-e4b-it-q4_k_m.gguf -c 4096
# For chat interface
./llama-chat -m gemma-3n-e4b-it-q4_k_m.gguf --color -i
```
### Using with Python (llama-cpp-python)
```python
from llama_cpp import Llama
# Load model
llm = Llama(
model_path="gemma-3n-e4b-it-q4_k_m.gguf",
n_ctx=4096,
n_threads=8,
verbose=False
)
# Generate response
response = llm.create_chat_completion(
messages=[
{"role": "user", "content": "Hello! Can you introduce yourself?"}
],
max_tokens=512,
temperature=0.7
)
print(response['choices'][0]['message']['content'])
```
## 💬 **Chat Template**
Gemma 3n uses the following chat template:
```
<bos><start_of_turn>user
{user_message}<end_of_turn>
<start_of_turn>model
{assistant_response}<end_of_turn>
```
### Recommended Parameters
```yaml
temperature: 0.7
top_p: 0.9
top_k: 40
repeat_penalty: 1.1
max_tokens: 4096
stop_tokens: ["<end_of_turn>"]
```
## 📋 **Model Details**
- **Base Model**: [google/gemma-3n-e4b-it](https://huggingface.co/google/gemma-3n-e4b-it)
- **Architecture**: Gemma 3n (Effective 4B parameters)
- **Context Length**: 32,768 tokens
- **Vocabulary Size**: 256,000 tokens
- **Training Data**: Multilingual (140+ languages)
- **License**: Gemma License
### Key Features
- ✅ **Efficient Architecture**: Uses selective parameter activation
- ✅ **Multilingual Support**: Trained on 140+ spoken languages
- ✅ **Instruction-Tuned**: Optimized for conversational AI
- ✅ **Extended Context**: 32K token context window
- ✅ **Hardware Optimized**: Designed for everyday devices
## 🔧 **Quantization Details**
All models were quantized using the latest **llama.cpp** with full Gemma 3n architecture support:
### Quantization Process
```bash
# Example quantization command
./llama-quantize input.gguf output.gguf q4_k_m
```
### Architecture-Specific Tensors
The quantization properly handles Gemma 3n-specific components:
- `altup_correct_coef`, `altup_correct_scale`, `altup_predict_coef`
- `altup_router`, `altup_router_norm`
- `laurel_l`, `laurel_r`, `laurel_post_norm`
- Standard attention, FFN, and normalization layers
## 📈 **Performance Benchmarks**
Based on the original Gemma 3n E4B IT model:
| **Benchmark** | **Score** | **Category** |
|---------------|-----------|--------------|
| MMLU | 64.9% | General Knowledge |
| HumanEval | 75.0% | Code Generation |
| HellaSwag | 78.6% | Commonsense Reasoning |
| ARC-E | 81.6% | Elementary Science |
| TriviaQA | 70.2% | Factual Knowledge |
| MBPP | 63.6% | Programming |
## 🎯 **Recommended Usage**
### For Different Hardware:
- **High-end GPU (24GB+ VRAM)**: Q8_0 or F16
- **Mid-range GPU (12-16GB VRAM)**: Q6_K or Q5_K_M
- **Consumer GPU (8-12GB VRAM)**: Q4_K_M or Q5_K_S
- **CPU-only or low VRAM**: Q4_0 or Q3_K_M
- **Mobile/Edge devices**: Q3_K_S or Q2_K
### For Different Use Cases:
- **Production chatbots**: Q5_K_M or Q4_K_M
- **Research/evaluation**: Q8_0 or F16
- **Rapid prototyping**: Q4_0
- **Resource-constrained**: Q3_K_M
- **Edge deployment**: Q2_K
## 🛠️ **Technical Requirements**
### Minimum System Requirements:
- **RAM**: 4GB (for Q2_K) to 16GB (for F16)
- **Storage**: 3GB to 13GB depending on quantization
- **CPU**: Modern x64 processor with AVX support
- **GPU** (optional): CUDA, OpenCL, or Metal support
### Software Dependencies:
- **llama.cpp**: Latest version with Gemma 3n support
- **Python**: 3.8+ for llama-cpp-python
- **Ollama**: 0.10.0+ for official support
## 📚 **Usage Examples**
### Example Conversations:
```
User: What is Gemma 3n and how does it differ from other models?
Gemma 3n: I'm Gemma 3n, a helpful AI assistant created by Google. I'm part of the Gemma 3n family, which is designed for efficient execution on everyday devices like laptops, tablets, and phones.
What makes me different from other models:
- **Selective Parameter Activation**: I use an innovative technology that allows me to operate effectively with 4B parameters while containing more total parameters
- **Multilingual Capabilities**: I was trained on data in over 140 spoken languages
- **Hardware Efficiency**: I'm optimized to run well on consumer hardware, not just high-end servers
- **Extended Context**: I can handle up to 32K tokens in context
- **Instruction-Tuned**: I'm specifically fine-tuned for conversational interactions
User: Can you help me write a Python function to calculate fibonacci numbers?
Gemma 3n: Certainly! Here are a few different approaches to calculate Fibonacci numbers:
def fibonacci_recursive(n):
"""Recursive approach - simple but inefficient for large n"""
if n <= 1:
return n
return fibonacci_recursive(n-1) + fibonacci_recursive(n-2)
def fibonacci_iterative(n):
"""Iterative approach - efficient and readable"""
if n <= 1:
return n
a, b = 0, 1
for _ in range(2, n + 1):
a, b = b, a + b
return b
def fibonacci_generator(limit):
"""Generator for Fibonacci sequence up to limit"""
a, b = 0, 1
while a < limit:
yield a
a, b = b, a + b
# Examples:
print(fibonacci_iterative(10)) # Output: 55
print(list(fibonacci_generator(100))) # All Fib numbers < 100
```
## 🔧 **Troubleshooting**
### Common Issues:
**1. "unknown model architecture" error:**
- Ensure you're using the latest llama.cpp with Gemma 3n support
- Update to llama.cpp commit `8846aace` or later
**2. Ollama compatibility issues:**
- Update Ollama to version 0.10.0 or later for native Gemma 3n support
- For older Ollama versions, use custom Modelfile approach
**3. Out of memory errors:**
- Try a smaller quantization (Q4_0, Q3_K_M, or Q2_K)
- Reduce context length with `-c` parameter
- Use CPU inference instead of GPU
**4. Slow inference:**
- Use GPU acceleration if available
- Try quantizations like Q4_K_M or Q4_0 for better speed
- Adjust thread count with `-t` parameter
### Performance Tips:
- **GPU users**: Q4_K_M or Q5_K_M offer the best speed/quality balance
- **CPU users**: Q4_0 provides good compatibility and reasonable speed
- **Memory-constrained**: Q3_K_M or Q2_K can run on limited hardware
## ⚠️ **Limitations**
- **Quantization Quality**: Lower bit quantizations (Q2_K, Q3_K) may have reduced quality
- **Context Length**: While supporting 32K tokens, performance may degrade with very long contexts
- **Language Coverage**: While multilingual, performance may vary across different languages
- **Knowledge Cutoff**: Training data has a cutoff date (June 2024)
- **Factual Accuracy**: May generate plausible but incorrect information
- **Code Generation**: While capable, may require verification for production use
## 🔒 **Ethical Considerations**
- **Bias**: Model may reflect biases present in training data
- **Safety**: Content filtering is not included in these GGUF versions
- **Verification**: Always verify outputs, especially for critical applications
- **Responsible Use**: Follow the Gemma License and ethical AI guidelines
## 📄 **Files Included**
```
gemma-3n-e4b-it-f16.gguf # 13.0 GB - Full precision
gemma-3n-e4b-it-q8_0.gguf # 6.8 GB - 8-bit quantization
gemma-3n-e4b-it-q6_k.gguf # 5.3 GB - 6-bit K-quant
gemma-3n-e4b-it-q5_1.gguf # 4.9 GB - 5-bit (high quality)
gemma-3n-e4b-it-q5_k_m.gguf # 4.6 GB - 5-bit K-quant medium
gemma-3n-e4b-it-q5_k_s.gguf # 4.5 GB - 5-bit K-quant small
gemma-3n-e4b-it-q5_0.gguf # 4.5 GB - 5-bit standard
gemma-3n-e4b-it-q4_1.gguf # 4.2 GB - 4-bit (improved)
gemma-3n-e4b-it-q4_k_m.gguf # 3.9 GB - 4-bit K-quant medium
gemma-3n-e4b-it-q4_k_s.gguf # 3.8 GB - 4-bit K-quant small
gemma-3n-e4b-it-q4_0.gguf # 3.8 GB - 4-bit standard
gemma-3n-e4b-it-q3_k_l.gguf # 3.4 GB - 3-bit K-quant large
gemma-3n-e4b-it-q3_k_m.gguf # 3.2 GB - 3-bit K-quant medium
gemma-3n-e4b-it-q3_k_s.gguf # 3.0 GB - 3-bit K-quant small
gemma-3n-e4b-it-q2_k.gguf # 2.6 GB - 2-bit K-quant
```
## 🙏 **Acknowledgments**
- **Google DeepMind**: For developing and releasing Gemma 3n
- **llama.cpp community**: For implementing Gemma 3n architecture support
- **Hugging Face**: For providing the model hosting platform
- **Quantization**: Performed using the latest llama.cpp tools
## 📞 **Support & Community**
- **Issues**: Report problems in the repository issues
- **Discussions**: Join the community discussions
- **Updates**: Follow for model updates and improvements
## 📜 **Citation**
If you use these models in your research, please cite:
```bibtex
@article{gemma_3n_2025,
title={Gemma 3n},
url={https://ai.google.dev/gemma/docs/gemma-3n},
publisher={Google DeepMind},
author={Gemma Team},
year={2025}
}
```
---
**License**: This model is released under the Gemma License. Please review the license terms before use.
**Disclaimer**: These quantized models are provided for research and educational purposes. Users are responsible for ensuring compliance with applicable laws and ethical guidelines.
# Gemma 3n model card
**Model Page**: [Gemma 3n](https://ai.google.dev/gemma/docs/gemma-3n)
**Resources and Technical Documentation**:
- [Responsible Generative AI Toolkit](https://ai.google.dev/responsible)
- [Gemma on Kaggle](https://www.kaggle.com/models/google/gemma-3n)
- [Gemma on HuggingFace](https://huggingface.co/collections/google/gemma-3n-685065323f5984ef315c93f4)
- [Gemma on Vertex Model Garden](https://console.cloud.google.com/vertex-ai/publishers/google/model-garden/gemma3n)
**Terms of Use**: [Terms](https://ai.google.dev/gemma/terms)\
**Authors**: Google DeepMind
## Model Information
Summary description and brief definition of inputs and outputs.
### Description
Gemma is a family of lightweight, state-of-the-art open models from Google,
built from the same research and technology used to create the Gemini models.
Gemma 3n models are designed for efficient execution on low-resource devices.
They are capable of multimodal input, handling text, image, video, and audio
input, and generating text outputs, with open weights for pre-trained and
instruction-tuned variants. These models were trained with data in over 140
spoken languages.
Gemma 3n models use selective parameter activation technology to reduce resource
requirements. This technique allows the models to operate at an effective size
of 2B and 4B parameters, which is lower than the total number of parameters they
contain. For more information on Gemma 3n's efficient parameter management
technology, see the
[Gemma 3n](https://ai.google.dev/gemma/docs/gemma-3n#parameters)
page.
### Inputs and outputs
- **Input:**
- Text string, such as a question, a prompt, or a document to be
summarized
- Images, normalized to 256x256, 512x512, or 768x768 resolution
and encoded to 256 tokens each
- Audio data encoded to 6.25 tokens per second from a single channel
- Total input context of 32K tokens
- **Output:**
- Generated text in response to the input, such as an answer to a
question, analysis of image content, or a summary of a document
- Total output length up to 32K tokens, subtracting the request
input tokens
### Usage
Below, there are some code snippets on how to get quickly started with running
the model. First, install the Transformers library. Gemma 3n is supported
starting from transformers 4.53.0.
```sh
$ pip install -U transformers
```
Then, copy the snippet from the section that is relevant for your use case.
#### Running with the `pipeline` API
You can initialize the model and processor for inference with `pipeline` as
follows.
```python
from transformers import pipeline
import torch
pipe = pipeline(
"image-text-to-text",
model="google/gemma-3n-e4b-it",
device="cuda",
torch_dtype=torch.bfloat16,
)
```
With instruction-tuned models, you need to use chat templates to process our
inputs first. Then, you can pass it to the pipeline.
```python
messages = [
{
"role": "system",
"content": [{"type": "text", "text": "You are a helpful assistant."}]
},
{
"role": "user",
"content": [
{"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
{"type": "text", "text": "What animal is on the candy?"}
]
}
]
output = pipe(text=messages, max_new_tokens=200)
print(output[0]["generated_text"][-1]["content"])
# Okay, let's take a look!
# Based on the image, the animal on the candy is a **turtle**.
# You can see the shell shape and the head and legs.
```
#### Running the model on a single GPU
```python
from transformers import AutoProcessor, Gemma3nForConditionalGeneration
from PIL import Image
import requests
import torch
model_id = "google/gemma-3n-e4b-it"
model = Gemma3nForConditionalGeneration.from_pretrained(model_id, device_map="auto", torch_dtype=torch.bfloat16,).eval()
processor = AutoProcessor.from_pretrained(model_id)
messages = [
{
"role": "system",
"content": [{"type": "text", "text": "You are a helpful assistant."}]
},
{
"role": "user",
"content": [
{"type": "image", "image": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg"},
{"type": "text", "text": "Describe this image in detail."}
]
}
]
inputs = processor.apply_chat_template(
messages,
add_generation_prompt=True,
tokenize=True,
return_dict=True,
return_tensors="pt",
).to(model.device)
input_len = inputs["input_ids"].shape[-1]
with torch.inference_mode():
generation = model.generate(**inputs, max_new_tokens=100, do_sample=False)
generation = generation[0][input_len:]
decoded = processor.decode(generation, skip_special_tokens=True)
print(decoded)
# **Overall Impression:** The image is a close-up shot of a vibrant garden scene,
# focusing on a cluster of pink cosmos flowers and a busy bumblebee.
# It has a slightly soft, natural feel, likely captured in daylight.
```
### Citation
```
@article{gemma_3n_2025,
title={Gemma 3n},
url={https://ai.google.dev/gemma/docs/gemma-3n},
publisher={Google DeepMind},
author={Gemma Team},
year={2025}
}
```
## Model Data
Data used for model training and how the data was processed.
### Training Dataset
These models were trained on a dataset that includes a wide variety of sources
totalling approximately 11 trillion tokens. The knowledge cutoff date for the
training data was June 2024. Here are the key components:
- **Web Documents**: A diverse collection of web text ensures the model
is exposed to a broad range of linguistic styles, topics, and vocabulary.
The training dataset includes content in over 140 languages.
- **Code**: Exposing the model to code helps it to learn the syntax and
patterns of programming languages, which improves its ability to generate
code and understand code-related questions.
- **Mathematics**: Training on mathematical text helps the model learn
logical reasoning, symbolic representation, and to address mathematical queries.
- **Images**: A wide range of images enables the model to perform image
analysis and visual data extraction tasks.
- Audio: A diverse set of sound samples enables the model to recognize
speech, transcribe text from recordings, and identify information in audio data.
The combination of these diverse data sources is crucial for training a
powerful multimodal model that can handle a wide variety of different tasks and
data formats.
### Data Preprocessing
Here are the key data cleaning and filtering methods applied to the training
data:
- **CSAM Filtering**: Rigorous CSAM (Child Sexual Abuse Material)
filtering was applied at multiple stages in the data preparation process to
ensure the exclusion of harmful and illegal content.
- **Sensitive Data Filtering**: As part of making Gemma pre-trained models
safe and reliable, automated techniques were used to filter out certain
personal information and other sensitive data from training sets.
- **Additional methods**: Filtering based on content quality and safety in
line with
[our policies](https://ai.google/static/documents/ai-responsibility-update-published-february-2025.pdf).
## Implementation Information
Details about the model internals.
### Hardware
Gemma was trained using [Tensor Processing Unit
(TPU)](https://cloud.google.com/tpu/docs/intro-to-tpu) hardware (TPUv4p, TPUv5p
and TPUv5e). Training generative models requires significant computational
power. TPUs, designed specifically for matrix operations common in machine
learning, offer several advantages in this domain:
- **Performance**: TPUs are specifically designed to handle the massive
computations involved in training generative models. They can speed up
training considerably compared to CPUs.
- **Memory**: TPUs often come with large amounts of high-bandwidth memory,
allowing for the handling of large models and batch sizes during training.
This can lead to better model quality.
- **Scalability**: TPU Pods (large clusters of TPUs) provide a scalable
solution for handling the growing complexity of large foundation models.
You can distribute training across multiple TPU devices for faster and more
efficient processing.
- **Cost-effectiveness**: In many scenarios, TPUs can provide a more
cost-effective solution for training large models compared to CPU-based
infrastructure, especially when considering the time and resources saved
due to faster training.
These advantages are aligned with
[Google's commitments to operate sustainably](https://sustainability.google/operating-sustainably/).
### Software
Training was done using [JAX](https://github.com/jax-ml/jax) and
[ML Pathways](https://blog.google/technology/ai/introducing-pathways-next-generation-ai-architecture/).
JAX allows researchers to take advantage of the latest generation of hardware,
including TPUs, for faster and more efficient training of large models. ML
Pathways is Google's latest effort to build artificially intelligent systems
capable of generalizing across multiple tasks. This is specially suitable for
foundation models, including large language models like these ones.
Together, JAX and ML Pathways are used as described in the
[paper about the Gemini family of models](https://goo.gle/gemma2report):
*"the 'single controller' programming model of Jax and Pathways allows a single
Python process to orchestrate the entire training run, dramatically simplifying
the development workflow."*
## Evaluation
Model evaluation metrics and results.
### Benchmark Results
These models were evaluated at full precision (float32) against a large
collection of different datasets and metrics to cover different aspects of
content generation. Evaluation results marked with **IT** are for
instruction-tuned models. Evaluation results marked with **PT** are for
pre-trained models.
#### Reasoning and factuality
| Benchmark | Metric | n-shot | E2B PT | E4B PT |
| ------------------------------ |----------------|----------|:--------:|:--------:|
| [HellaSwag][hellaswag] | Accuracy | 10-shot | 72.2 | 78.6 |
| [BoolQ][boolq] | Accuracy | 0-shot | 76.4 | 81.6 |
| [PIQA][piqa] | Accuracy | 0-shot | 78.9 | 81.0 |
| [SocialIQA][socialiqa] | Accuracy | 0-shot | 48.8 | 50.0 |
| [TriviaQA][triviaqa] | Accuracy | 5-shot | 60.8 | 70.2 |
| [Natural Questions][naturalq] | Accuracy | 5-shot | 15.5 | 20.9 |
| [ARC-c][arc] | Accuracy | 25-shot | 51.7 | 61.6 |
| [ARC-e][arc] | Accuracy | 0-shot | 75.8 | 81.6 |
| [WinoGrande][winogrande] | Accuracy | 5-shot | 66.8 | 71.7 |
| [BIG-Bench Hard][bbh] | Accuracy | few-shot | 44.3 | 52.9 |
| [DROP][drop] | Token F1 score | 1-shot | 53.9 | 60.8 |
[hellaswag]: https://arxiv.org/abs/1905.07830
[boolq]: https://arxiv.org/abs/1905.10044
[piqa]: https://arxiv.org/abs/1911.11641
[socialiqa]: https://arxiv.org/abs/1904.09728
[triviaqa]: https://arxiv.org/abs/1705.03551
[naturalq]: https://github.com/google-research-datasets/natural-questions
[arc]: https://arxiv.org/abs/1911.01547
[winogrande]: https://arxiv.org/abs/1907.10641
[bbh]: https://paperswithcode.com/dataset/bbh
[drop]: https://arxiv.org/abs/1903.00161
#### Multilingual
| Benchmark | Metric | n-shot | E2B IT | E4B IT |
| ------------------------------------|-------------------------|----------|:--------:|:--------:|
| [MGSM][mgsm] | Accuracy | 0-shot | 53.1 | 60.7 |
| [WMT24++][wmt24pp] (ChrF) | Character-level F-score | 0-shot | 42.7 | 50.1 |
| [Include][include] | Accuracy | 0-shot | 38.6 | 57.2 |
| [MMLU][mmlu] (ProX) | Accuracy | 0-shot | 8.1 | 19.9 |
| [OpenAI MMLU][openai-mmlu] | Accuracy | 0-shot | 22.3 | 35.6 |
| [Global-MMLU][global-mmlu] | Accuracy | 0-shot | 55.1 | 60.3 |
| [ECLeKTic][eclektic] | ECLeKTic score | 0-shot | 2.5 | 1.9 |
[mgsm]: https://arxiv.org/abs/2210.03057
[wmt24pp]: https://arxiv.org/abs/2502.12404v1
[include]:https://arxiv.org/abs/2411.19799
[mmlu]: https://arxiv.org/abs/2009.03300
[openai-mmlu]: https://huggingface.co/datasets/openai/MMMLU
[global-mmlu]: https://huggingface.co/datasets/CohereLabs/Global-MMLU
[eclektic]: https://arxiv.org/abs/2502.21228
#### STEM and code
| Benchmark | Metric | n-shot | E2B IT | E4B IT |
| ------------------------------------|--------------------------|----------|:--------:|:--------:|
| [GPQA][gpqa] Diamond | RelaxedAccuracy/accuracy | 0-shot | 24.8 | 23.7 |
| [LiveCodeBench][lcb] v5 | pass@1 | 0-shot | 18.6 | 25.7 |
| Codegolf v2.2 | pass@1 | 0-shot | 11.0 | 16.8 |
| [AIME 2025][aime-2025] | Accuracy | 0-shot | 6.7 | 11.6 |
[gpqa]: https://arxiv.org/abs/2311.12022
[lcb]: https://arxiv.org/abs/2403.07974
[aime-2025]: https://www.vals.ai/benchmarks/aime-2025-05-09
#### Additional benchmarks
| Benchmark | Metric | n-shot | E2B IT | E4B IT |
| ------------------------------------ |------------|----------|:--------:|:--------:|
| [MMLU][mmlu] | Accuracy | 0-shot | 60.1 | 64.9 |
| [MBPP][mbpp] | pass@1 | 3-shot | 56.6 | 63.6 |
| [HumanEval][humaneval] | pass@1 | 0-shot | 66.5 | 75.0 |
| [LiveCodeBench][lcb] | pass@1 | 0-shot | 13.2 | 13.2 |
| HiddenMath | Accuracy | 0-shot | 27.7 | 37.7 |
| [Global-MMLU-Lite][global-mmlu-lite] | Accuracy | 0-shot | 59.0 | 64.5 |
| [MMLU][mmlu] (Pro) | Accuracy | 0-shot | 40.5 | 50.6 |
[gpqa]: https://arxiv.org/abs/2311.12022
[mbpp]: https://arxiv.org/abs/2108.07732
[humaneval]: https://arxiv.org/abs/2107.03374
[lcb]: https://arxiv.org/abs/2403.07974
[global-mmlu-lite]: https://huggingface.co/datasets/CohereForAI/Global-MMLU-Lite
## Ethics and Safety
Ethics and safety evaluation approach and results.
### Evaluation Approach
Our evaluation methods include structured evaluations and internal red-teaming
testing of relevant content policies. Red-teaming was conducted by a number of
different teams, each with different goals and human evaluation metrics. These
models were evaluated against a number of different categories relevant to
ethics and safety, including:
- **Child Safety**: Evaluation of text-to-text and image to text prompts
covering child safety policies, including child sexual abuse and
exploitation.
- **Content Safety:** Evaluation of text-to-text and image to text prompts
covering safety policies including, harassment, violence and gore, and hate
speech.
- **Representational Harms**: Evaluation of text-to-text and image to text
prompts covering safety policies including bias, stereotyping, and harmful
associations or inaccuracies.
In addition to development level evaluations, we conduct "assurance
evaluations" which are our 'arms-length' internal evaluations for responsibility
governance decision making. They are conducted separately from the model
development team, to inform decision making about release. High level findings
are fed back to the model team, but prompt sets are held-out to prevent
overfitting and preserve the results' ability to inform decision making. Notable
assurance evaluation results are reported to our Responsibility & Safety Council
as part of release review.
### Evaluation Results
For all areas of safety testing, we saw safe levels of performance across the
categories of child safety, content safety, and representational harms relative
to previous Gemma models. All testing was conducted without safety filters to
evaluate the model capabilities and behaviors. For text-to-text, image-to-text,
and audio-to-text, and across all model sizes, the model produced minimal policy
violations, and showed significant improvements over previous Gemma models'
performance with respect to high severity violations. A limitation of our
evaluations was they included primarily English language prompts.
## Usage and Limitations
These models have certain limitations that users should be aware of.
### Intended Usage
Open generative models have a wide range of applications across various
industries and domains. The following list of potential uses is not
comprehensive. The purpose of this list is to provide contextual information
about the possible use-cases that the model creators considered as part of model
training and development.
- Content Creation and Communication
- **Text Generation**: Generate creative text formats such as
poems, scripts, code, marketing copy, and email drafts.
- **Chatbots and Conversational AI**: Power conversational
interfaces for customer service, virtual assistants, or interactive
applications.
- **Text Summarization**: Generate concise summaries of a text
corpus, research papers, or reports.
- **Image Data Extraction**: Extract, interpret, and summarize
visual data for text communications.
- **Audio Data Extraction**: Transcribe spoken language, translate speech
to text in other languages, and analyze sound-based data.
- Research and Education
- **Natural Language Processing (NLP) and generative model
Research**: These models can serve as a foundation for researchers to
experiment with generative models and NLP techniques, develop
algorithms, and contribute to the advancement of the field.
- **Language Learning Tools**: Support interactive language
learning experiences, aiding in grammar correction or providing writing
practice.
- **Knowledge Exploration**: Assist researchers in exploring large
bodies of data by generating summaries or answering questions about
specific topics.
### Limitations
- Training Data
- The quality and diversity of the training data significantly
influence the model's capabilities. Biases or gaps in the training data
can lead to limitations in the model's responses.
- The scope of the training dataset determines the subject areas
the model can handle effectively.
- Context and Task Complexity
- Models are better at tasks that can be framed with clear
prompts and instructions. Open-ended or highly complex tasks might be
challenging.
- A model's performance can be influenced by the amount of context
provided (longer context generally leads to better outputs, up to a
certain point).
- Language Ambiguity and Nuance
- Natural language is inherently complex. Models might struggle
to grasp subtle nuances, sarcasm, or figurative language.
- Factual Accuracy
- Models generate responses based on information they learned
from their training datasets, but they are not knowledge bases. They
may generate incorrect or outdated factual statements.
- Common Sense
- Models rely on statistical patterns in language. They might
lack the ability to apply common sense reasoning in certain situations.
### Ethical Considerations and Risks
The development of generative models raises several ethical concerns. In
creating an open model, we have carefully considered the following:
- Bias and Fairness
- Generative models trained on large-scale, real-world text and image data
can reflect socio-cultural biases embedded in the training material.
These models underwent careful scrutiny, input data pre-processing
described and posterior evaluations reported in this card.
- Misinformation and Misuse
- Generative models can be misused to generate text that is
false, misleading, or harmful.
- Guidelines are provided for responsible use with the model, see the
[Responsible Generative AI Toolkit](https://ai.google.dev/responsible).
- Transparency and Accountability:
- This model card summarizes details on the models' architecture,
capabilities, limitations, and evaluation processes.
- A responsibly developed open model offers the opportunity to
share innovation by making generative model technology accessible to
developers and researchers across the AI ecosystem.
Risks identified and mitigations:
- **Perpetuation of biases**: It's encouraged to perform continuous monitoring
(using evaluation metrics, human review) and the exploration of de-biasing
techniques during model training, fine-tuning, and other use cases.
- **Generation of harmful content**: Mechanisms and guidelines for content
safety are essential. Developers are encouraged to exercise caution and
implement appropriate content safety safeguards based on their specific
product policies and application use cases.
- **Misuse for malicious purposes**: Technical limitations and developer
and end-user education can help mitigate against malicious applications of
generative models. Educational resources and reporting mechanisms for users
to flag misuse are provided. Prohibited uses of Gemma models are outlined
in the
[Gemma Prohibited Use Policy](https://ai.google.dev/gemma/prohibited_use_policy).
- **Privacy violations**: Models were trained on data filtered for removal of
certain personal information and other sensitive data. Developers are
encouraged to adhere to privacy regulations with privacy-preserving
techniques.
### Benefits
At the time of release, this family of models provides high-performance open
generative model implementations designed from the ground up for responsible AI
development compared to similarly sized models.
Using the benchmark evaluation metrics described in this document, these models
have shown to provide superior performance to other, comparably-sized open model
alternatives.

3
gemma-3n-e4b-it-f16.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:34de93a7c3c12ce0032110e35e2c05aac89f1264abab384fefbc9146749fabd7
size 13740103424

3
gemma-3n-e4b-it-q2_k.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:a42e301e63dee68086acb546dff31b58b6fde2c48375885a036004edce42a6c9
size 2757318400

3
gemma-3n-e4b-it-q3_k_l.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:8db57a9288c0698c09d32336a04dd91722c4ef5591436ed1531c79aec0f3e1b0
size 3601995520

3
gemma-3n-e4b-it-q3_k_m.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:0082a72e4f4a1ed10cd0bc21abc30b313f803b7be185379dc57737977847ce70
size 3440908032

3
gemma-3n-e4b-it-q3_k_s.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:3d403f233f10c50f101a3893d72a1399524f9c31c07641eed07352b836bdf320
size 3251910400

3
gemma-3n-e4b-it-q4_0.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:1a418ad202f79d59e4750686dc2bd0ab5080e8cb28377d7f7d2dcccd124d7560
size 4085405440

3
gemma-3n-e4b-it-q4_1.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:8e6427d9abaff347a6f0035d8d5a97b324881369d34bfce617efa8df67d3aef3
size 4477638400

3
gemma-3n-e4b-it-q4_k_m.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:8787ebd821cfcfeabd46a3a629747747d0df8cdbf8df00d8b5bf6f3b0aa37564
size 4237063936

3
gemma-3n-e4b-it-q4_k_s.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:ed137264f2d9c882d209f47c6adfba87802f43ee1781c9aee1a2764564ec852a
size 4102706944

3
gemma-3n-e4b-it-q5_0.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:def2a94ecc699a05fc433068ddb14ae85162e6df8c76f5341e4cda75f3415022
size 4869871360

3
gemma-3n-e4b-it-q5_1.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:166c352191af1df930d31977ca6b6cb8e75868f69c7b6eb91a853d6cf39209e6
size 5262104320

3
gemma-3n-e4b-it-q5_k_m.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:5f934154eee9fd8c2580d87d10c3a8e6f4e5b8821452680f8a58d6c678bf34d1
size 4947998464

3
gemma-3n-e4b-it-q5_k_s.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:5e27e4cfe3a8a5586b18f25f2b9918211f2d741f74ef7679878604b0f0141d33
size 4869871360

3
gemma-3n-e4b-it-q6_k.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:fe210f251e725916fd50d8bd39c10a54d9226266ef6dc7c324f85f16b1a5d894
size 5703366400

3
gemma-3n-e4b-it-q8_0.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:f2c5e98a7564d73f23019c015ed6b4a44a2ebd6780e55205ead7f9eb6b95e72d
size 7353292544

11
params Normal file
View File

@@ -0,0 +1,11 @@
{
"stop": [
"<end_of_turn>",
"<eos>"
],
"temperature": 0.6,
"min_p" : 0.00,
"repeat_penalty" : 1.0,
"top_k" : 64,
"top_p" : 0.95
}

11
template Normal file
View File

@@ -0,0 +1,11 @@
{{- range $i, $_ := .Messages }}
{{- $last := eq (len (slice $.Messages $i)) 1 }}
{{- if or (eq .Role "user") (eq .Role "system") }}<start_of_turn>user
{{ .Content }}<end_of_turn>
{{ if $last }}<start_of_turn>model
{{ end }}
{{- else if eq .Role "assistant" }}<start_of_turn>model
{{ .Content }}{{ if not $last }}<end_of_turn>
{{ end }}
{{- end }}
{{- end }}