add separate quantization primitives for float8

ghstack-source-id: bc601f2
ghstack-comment-id: 2608048970
Pull Request resolved: #1597

Author: danielvegamyhre

test/quantization/test_quant_primitives.py

@@ -9,16 +9,21 @@
 import unittest
 
 import torch
+from parameterized import parameterized
 
 from torchao.dtypes.utils import is_device
+from torchao.float8.float8_utils import EPS as float8_eps
 from torchao.quantization.quant_primitives import (
-    MappingType,
-    ZeroPointDomain,
     choose_qparams_affine,
+    choose_qparams_affine_float8,
     dequantize_affine,
+    dequantize_affine_float8,
     fake_quantize_affine,
     fake_quantize_affine_cachemask,
+    MappingType,
     quantize_affine,
+    quantize_affine_float8,
+    ZeroPointDomain,
 )
 
 # TODO: remove test for utils?
@@ -29,11 +34,11 @@
     quantize_activation_per_token_absmax,
 )
 from torchao.utils import (
+    is_fbcode,
     TORCH_VERSION_AT_LEAST_2_3,
     TORCH_VERSION_AT_LEAST_2_4,
     TORCH_VERSION_AT_LEAST_2_5,
     TORCH_VERSION_AT_LEAST_2_6,
-    is_fbcode,
 )
 
 _SEED = 1234
@@ -838,6 +843,58 @@ def test_fake_quantize_affine_cachemask(self):
         torch.testing.assert_close(dequantized, fake_quantized)
         torch.testing.assert_close(expected_mask, mask)
 
+    @parameterized.expand(
+        [
+            (
+                torch.float32,
+                torch.float8_e4m3fn,
+            ),
+        ]
+    )
+    def test_float8_quant_primitives(self, hp_dtype, float8_dtype):
+        input = torch.randn(10, 10)
+
+        # float8 quantization primitives
+        scale = choose_qparams_affine_float8(input, float8_dtype)
+        quantized = quantize_affine_float8(input, scale, float8_dtype)
+        dequantized = dequantize_affine_float8(quantized, scale, hp_dtype)
+
+        # reference implementation using generic primitives
+        expected_scale, _ = choose_qparams_affine(
+            input,
+            MappingType.SYMMETRIC,
+            input.shape,
+            float8_dtype,
+            eps=float8_eps,  # use same EPS as float8 training
+            scale_dtype=torch.float32,
+            quant_min=torch.finfo(float8_dtype).min,
+            quant_max=torch.finfo(float8_dtype).max,
+        )
+        expected_quantized = quantize_affine(
+            input,
+            input.shape,
+            scale,
+            output_dtype=float8_dtype,
+            quant_min=torch.finfo(float8_dtype).min,
+            quant_max=torch.finfo(float8_dtype).max,
+            zero_point=None,
+            zero_point_domain=None,
+        )
+        expected_dequantized = dequantize_affine(
+            expected_quantized,
+            input.shape,
+            scale,
+            input_dtype=float8_dtype,
+            quant_min=torch.finfo(float8_dtype).min,
+            quant_max=torch.finfo(float8_dtype).max,
+            zero_point=None,
+            zero_point_domain=None,
+        )
+
+        self.assertTrue(torch.equal(expected_scale, scale))
+        torch.testing.assert_close(expected_quantized, quantized)
+        torch.testing.assert_close(expected_dequantized, dequantized)
+
 
 if __name__ == "__main__":
     unittest.main()

torchao/quantization/quant_primitives.py

@@ -39,6 +39,9 @@
     "MappingType",
     "ZeroPointDomain",
     "TorchAODType",
+    "choose_qparams_affine_float8",
+    "quantize_affine_float8",
+    "dequantize_affine_float8",
 ]
 
 
@@ -1300,3 +1303,69 @@ def dequantize_affine_floatx(
     tensor = tensor * scale.float().view(-1, 1)
     tensor = tensor.to(dtype=output_dtype)
     return tensor
+
+
+def choose_qparams_affine_float8(
+    tensor: torch.Tensor, float8_dtype: torch.dtype
+) -> torch.Tensor:
+    """
+    Calculates float8 scaling factor for the given high precision tensor, using tensorwise granularity.
+
+    Args:
+        tensor (torch.Tensor): Input tensor to be quantized.
+        float8_dtype (torch.dtype): Data type of the quantized tensor (e.g., torch.float8_e4m3fn, torch.float8_e5m2).
+    """
+    # only tensorwise scaling is supported for now:
+    quant_min, quant_max = torch.finfo(float8_dtype).min, torch.finfo(float8_dtype).max
+    min_val_neg = torch.min(tensor)
+    max_val_pos = torch.max(tensor)
+    max_val_pos = torch.max(-min_val_neg, max_val_pos)
+    scale = max_val_pos / (float(quant_max - quant_min) / 2)
+    return scale.to(dtype=torch.float32)
+
+    # max_val_pos = torch.max(-min_val_neg, max_val_pos)
+    # scale = max_val_pos / (float(quant_max - quant_min) / 2)
+
+
+def quantize_affine_float8(
+    tensor: torch.Tensor,
+    scale: torch.Tensor,
+    float8_dtype: torch.dtype,
+) -> torch.Tensor:
+    """
+    Quantizes the high precision floating point tensor to a float8 tensor, using the given scaling factor.
+
+    Args:
+        tensor (torch.Tensor): Input tensor to be quantized.
+        scale (torch.Tensor): Scaling factor for the quantization.
+        float8_dtype (torch.dtype): Data type of the quantized tensor (e.g., torch.float8_e4m3fn, torch.float8_e5m2).
+    """
+    # Note: when the line below is compiled with `torch.compile`, `tensor` is automatically
+    # upcasted to `float32` to multiply with the scale, since scale is a fp32 tensor in float8 quantization.
+    # In order to match numerics between eager and compile, we upcast manually here.
+    tensor_scaled = tensor.to(torch.float32) / scale
+    max_value = torch.finfo(float8_dtype).max
+    tensor_clamped = tensor_scaled.clamp(min=-max_value, max=max_value)
+    fp8_tensor = tensor_clamped.to(float8_dtype)
+    return fp8_tensor
+
+
+def dequantize_affine_float8(
+    tensor: torch.Tensor,
+    scale: torch.Tensor,
+    output_dtype: torch.dtype = torch.float32,
+) -> torch.Tensor:
+    """
+    Dequantizes the float8 tensor to high precision tensor.
+
+    Args:
+        tensor (torch.Tensor): Input float8 tensor to be dequantized.
+        scale (torch.Tensor): Scaling factor for the dequantization.
+        output_dtype (torch.dtype): Data type of the output tensor (e.g., torch.float32).
+    """
+    # Note: when the line below is compiled with `torch.compile`, `tensor` is automatically
+    # upcasted to `float32` to divide by the scale, since scale is a fp32 for float8 quantization.
+    # In order to match numerics between eager and compile, we upcast manually here.
+    fp8_tensor = tensor.to(torch.float32)
+    hp_tensor = fp8_tensor * scale
+    return hp_tensor.to(output_dtype)