---
language:
- multilingual
license: apache-2.0
---

# Model Card for Sindibad-7B



#  Table of Contents

0. [TL;DR](#TL;DR)
1. [Model Details](#model-details)
2. [Usage](#usage)
3. [Training Details](#training-details)
4. [Evaluation](#evaluation)


# TL;DR

# Model Details

## Model Description


- **Model type:** Language model
- **Language(s) (NLP):** English
- **License:** Apache 2.0

# Usage

Find below some example scripts on how to use the model in `transformers` (Make sure to have the latest transformers, or the one built from source):

## Using the Pytorch model

### Running the model on a CPU

<details>
<summary> Click to expand </summary>

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tiiuae/sindibad-7b")
model = AutoModelForCausalLM.from_pretrained("tiiuae/sindibad-7b")

input_text = "Question: How many hours in one day? Answer: "
input_ids = tokenizer(input_text, return_tensors="pt").input_ids

outputs = model.generate(input_ids)
print(tokenizer.decode(outputs[0]))
```

</details>

### Running the model on a GPU

<details>
<summary> Click to expand </summary>

```python
# pip install accelerate
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tiiuae/sindibad-7b")
model = AutoModelForCausalLM.from_pretrained("tiiuae/sindibad-7b", device_map="auto")

input_text = "Question: How many hours in one day? Answer: "
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")

outputs = model.generate(input_ids)
print(tokenizer.decode(outputs[0]))
```

</details>

### Running the model on a GPU using different precisions

#### FP16

<details>
<summary> Click to expand </summary>

```python
# pip install accelerate
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tiiuae/sindibad-7b")
model = AutoModelForCausalLM.from_pretrained("tiiuae/sindibad-7b", device_map="auto", torch_dtype=torch.float16)

input_text = "Question: How many hours in one day? Answer: "
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")

outputs = model.generate(input_ids)
print(tokenizer.decode(outputs[0]))
```

</details>

#### 4-bit

<details>
<summary> Click to expand </summary>

```python
# pip install bitsandbytes accelerate
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig

tokenizer = AutoTokenizer.from_pretrained("tiiuae/sindibad-7b")
model = AutoModelForCausalLM.from_pretrained("tiiuae/sindibad-7b", device_map="auto", quantization_config=BitsAndBytesConfig(load_in_4bit=True))

input_text = "Question: How many hours in one day? Answer: "
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")

outputs = model.generate(input_ids)
print(tokenizer.decode(outputs[0]))
```

</details>


# Training Details

Jingwei

## Training Data

Guillaume

## Training Procedure

The model was trained AdamW optimizer, WSD (warmup-stable-decay) learning rate schedule, and a batch size rampup from &nbsp; \\(b_{\mathrm{min}}=128\times2048\\)  &nbsp;  to &nbsp; \\(b_{\mathrm{max}}=2048\times2048\\) &nbsp; tokens during the first 50 GT of the training. In the stable phase, we used maximal learning rate &nbsp; \\(\eta_{\mathrm{max}}=6.4 \times 10^{-4}\\) &nbsp; and decayed it to the minimal value &nbsp; \\(\eta_{\mathrm{min}}=\eta_{\mathrm{max}} / 256\\) &nbsp; with exponential schedule over 500 GT. Also, we applied *BatchScaling* during the rampup — rescaling learning rate &nbsp; \\(\eta\\) &nbsp;so that the Adam noise temperature &nbsp; \\(T_{\mathrm{noise}}\equiv\eta / \sqrt{b}\\) &nbsp; is kept cosntant.  

# Evaluation

## Benchmarks

We evaluate our model on all benchmarks of the leaderboard's version 2 using the `lm-evaluation-harness` package, and we evaluate it on the benchmarks of version 1 using `lighteval`.

| model_name                   | IFEval | BBH   | MATH LvL5 | GPQA  | MUSR  | MMLU-PRO | **Average L2** | ARC   | HellaSwag | MMLU  | Winogrande | TruthfulQA | GSM8K | **Average L1** |
|------------------------------|--------|-------|-----------|-------|-------|----------|----------------|-------|-----------|-------|------------|------------|-------|----------------|
| `meta-llama/Meta-Llama-3-8B` | 14.55  | 24.50 | 3.25      | 7.38  | 6.24  | 24.55    | 13.41          | 60.24 | 82.23     | 66.70 | 78.45      | 42.93      | 45.19 | 62.62          |
| `tiiuae/falcon2-11B`         | 32.61  | 21.94 | 2.34      | 2.8   | 7.53  | 15.44    | 13.78          | 59.73 | 82.91     | 58.37 | 78.30      | 52.56      | 53.83 | **64.28**      |
| `mistralai/Mistral-7B-v0.1`  | 23.86  | 22.02 | 2.49      | 5.59  | 10.68 | 22.36    | 14.50          | 59.98 | 83.31     | 64.16 | 78.37      | 42.15      | 37.83 | 60.97          |
| `Zyphra/Zamba-7B-v1`         | -      | -     | -         | -     | -     | -        | -              | 46.48 | 80.24     | 57.72 | 76.4       | -          | -     | -              |
| Ours                         | 32.16  | 21.07 | 4.08      | 10.18 | 6.97  | 13.43    | **14.65**      | 61.69 | 80.63     | 61.05 | 74.03      | 53.60      | 51.86 | 63.81          |

## Throughput

This model can achieve comparable throughput and performance compared to other transformer based models that use optimized kernels such as Flash Attention 2. Make sure to install the optimized Mamba kernels with the following commands:

```bash
pip install "causal-conv1d>=1.4.0" mamba-ssm
```

Refer to our technical report for more details about performance evaluation.