Add example to README

Author: MarioSieg

README.md

@@ -37,7 +37,33 @@ To install pi-quant from PyPI, run the following command:
 pip install pypiquant
 ```
 
-# Benchmarks
+## Examples
+piquant is torch compatible. Here are some examples of how to use it with PyTorch:
+
+```python
+import torch
+import piquant
+
+# Quantize and back with: bfloat16 -> uint4 -> bfloat16
+# In torch, quint4x2 means two 4-bit quantized integers per byte.
+tensor = torch.rand(1000, dtype=torch.bfloat16, device='cpu')
+
+# Compute quantization parameters for uint4 (needed for quantization and dequantization)
+scale, zero_point = piquant.torch.compute_quant_params(tensor, dtype=torch.quint4x2)
+
+# Quantize the tensor to uint4
+quantized = piquant.torch.quantize(tensor, scale=scale, zero_point=zero_point, dtype=torch.quint4x2)
+
+# Dequantize back to bfloat16
+dequantized = piquant.torch.dequantize(quantized, scale=scale, zero_point=zero_point, dtype=torch.bfloat16)
+
+# Check if the dequantized tensor is close to the original tensor
+assert torch.allclose(dequantized, tensor, atol=scale*0.5 + 1e-3), "Dequantization did not match original tensor"
+
+# Print parts of original and dequantized tensors for verification
+print("Original tensor (first 10 elements):", tensor[:10].tolist())
+print("Dequant  tensor (first 10 elements):", dequantized[:10].tolist())
+```
 
 ## Benchmark