Finance Starter Pack

The Gradient Finance Starter Pack can be used to support foundational LLM development, adding additional insight into financial concepts, regulations, investments, and more. The training data is pre-cleaned and is designed to accelerate your foundational model on domain-specific knowledge in finance - supporting investors, insurers, and banking institutions. Use the training data to help build:

- Finance LLMs From Scratch
- Finance Co-Pilots
- AI Workflows Designed for Finance
- Etc.

Gradient is a leading innovator in building financial models, including 1) Albatross, Gradient's proprietary domain-specific model for financial services and 2) Alphatross, an earlier version of Gradient's Albatross model with limited capabilities - made available on Hugging Face.

Dataset examples:

```python
# Here we plot a reliablity diagram with custom bins (10 equally spaced).
# Since we have only 100 data points it is hard to see much of a pattern
# except that most points seem above the line.

plt.figure(figsize=(10,4))
mli.plot_reliability_diagram(yvec,xvec, bins=np.linspace(0,1,11), show_histogram=True);
```


```python
# Try SplineCalib with the default settings
sc4 = mli.SplineCalib()
sc4.fit(xvec, yvec)
```

```python
# Here is the resulting calibration curve
sc4.show_calibration_curve()
```


```python
# The curve does not go to [0,0], let's investigate...

sc4.calibrate(np.array([0,1e-16,1e-10,1e-9,1e-8,1e-7,1e-6,1e-5,1e-4,1e-3,1e-2]))
```
array([0.11966725, 0.11966725, 0.11966725, 0.11966725, 0.11966725,
0.11966725, 0.11966725, 0.11966725, 0.11966725, 0.11966725,
0.20690302])

```python
sc4.logodds_eps, np.min(xvec)
```

(0.001, 0.004695476192547066)

Dataset:
(text)

搜集汇总

背景与挑战

背景概述

该数据集为金融领域LLM开发提供预清洗的训练数据，涵盖金融概念、法规和投资等知识，支持构建金融专用模型和AI工作流。包含Gradient公司开发的Albatross等金融领域专业模型资源。