xref: /aosp_15_r20/external/executorch/kernels/quantized/quantized.yaml (revision 523fa7a60841cd1ecfb9cc4201f1ca8b03ed023a)
1- func: quantized_decomposed::add.out(Tensor a, float a_scale, int a_zero_point, int a_quant_min, int a_quant_max, Tensor b, float b_scale, int b_zero_point, int b_quant_min, int b_quant_max, float out_scale, int out_zero_point, int out_quant_min, int out_quant_max, *, Tensor(a!) out) -> Tensor(a!)
2  variants: function
3  kernels:
4    - arg_meta: null
5      kernel_name: torch::executor::quantized_add_out
6
7- func: quantized_decomposed::choose_qparams.Tensor_out(Tensor input, int quant_min, int quant_max, float eps, ScalarType dtype, *, Tensor(a!) scale_out, Tensor(b!) zero_point_out) -> (Tensor(a!), Tensor(b!))
8  variants: function
9  kernels:
10    - arg_meta: null
11      kernel_name: torch::executor::choose_qparams_tensor_out
12
13- func: quantized_decomposed::dequantize_per_tensor.out(Tensor input, float scale, int zero_point, int quant_min, int quant_max, ScalarType dtype, *, ScalarType? out_dtype=None, Tensor(a!) out) -> Tensor(a!)
14  variants: function
15  kernels:
16    - arg_meta: null
17      kernel_name: torch::executor::dequantize_per_tensor_out
18
19- func: quantized_decomposed::dequantize_per_tensor.Tensor_out(Tensor input, Tensor scale, Tensor zero_point, int quant_min, int quant_max, ScalarType dtype, *, ScalarType? out_dtype=None, Tensor(a!) out) -> Tensor(a!)
20  variants: function
21  kernels:
22    - arg_meta: null
23      kernel_name: torch::executor::dequantize_per_tensor_tensor_args_out
24
25- func: quantized_decomposed::quantize_per_channel.out(Tensor input, Tensor scales, Tensor zero_points, int axis, int quant_min, int quant_max, ScalarType dtype, *, Tensor(a!) out) -> Tensor(a!)
26  variants: function
27  kernels:
28    - arg_meta: null
29      kernel_name: torch::executor::quantize_per_channel_out
30
31- func: quantized_decomposed::dequantize_per_channel.out(Tensor input, Tensor scales, Tensor? zero_points, int axis, int quant_min, int quant_max, ScalarType dtype, *, ScalarType? out_dtype=None, Tensor(a!) out) -> Tensor(a!)
32  variants: function
33  kernels:
34    - arg_meta: null
35      kernel_name: torch::executor::dequantize_per_channel_out
36
37- func: quantized_decomposed::embedding_byte.out(Tensor weight, Tensor weight_scales, Tensor? weight_zero_points, int weight_quant_min, int weight_quant_max, Tensor indices, *, Tensor(a!) out) -> Tensor(a!)
38  variants: function
39  kernels:
40    - arg_meta: null
41      kernel_name: torch::executor::quantized_embedding_byte_out
42
43- func: quantized_decomposed::embedding_byte.dtype_out(Tensor weight, Tensor weight_scales, Tensor? weight_zero_points, int weight_quant_min, int weight_quant_max, Tensor indices, *, ScalarType? dtype=None, Tensor(a!) out) -> Tensor(a!)
44  variants: function
45  kernels:
46    - arg_meta: null
47      kernel_name: torch::executor::quantized_embedding_byte_dtype_out
48
49- func: quantized_decomposed::embedding_2bit.out(Tensor weight, Tensor weight_scales, Tensor? weight_zero_points, int weight_quant_min, int weight_quant_max, Tensor indices, *, Tensor(a!) out) -> Tensor(a!)
50  variants: function
51  kernels:
52    - arg_meta: null
53      kernel_name: torch::executor::quantized_embedding_2bit_out
54
55- func: quantized_decomposed::embedding_2bit.dtype_out(Tensor weight, Tensor weight_scales, Tensor? weight_zero_points, int weight_quant_min, int weight_quant_max, Tensor indices, ScalarType? dtype=None, *, Tensor(a!) out) -> Tensor(a!)
56  variants: function
57  kernels:
58    - arg_meta: null
59      kernel_name: torch::executor::quantized_embedding_2bit_dtype_out
60
61- func: quantized_decomposed::embedding_4bit.out(Tensor weight, Tensor weight_scales, Tensor? weight_zero_points, int weight_quant_min, int weight_quant_max, Tensor indices, *, Tensor(a!) out) -> Tensor(a!)
62  variants: function
63  kernels:
64    - arg_meta: null
65      kernel_name: torch::executor::quantized_embedding_4bit_out
66
67- func: quantized_decomposed::embedding_4bit.dtype_out(Tensor weight, Tensor weight_scales, Tensor? weight_zero_points, int weight_quant_min, int weight_quant_max, Tensor indices, ScalarType? dtype=None, *, Tensor(a!) out) -> Tensor(a!)
68  variants: function
69  kernels:
70    - arg_meta: null
71      kernel_name: torch::executor::quantized_embedding_4bit_dtype_out
72
73- func: quantized_decomposed::mixed_mm.out(Tensor input, Tensor weight, Tensor weight_scales, Tensor? weight_zero_points, *, Tensor(a!) out) -> Tensor(a!)
74  variants: function
75  kernels:
76    - arg_meta: null
77      kernel_name: torch::executor::quantized_mixed_mm_out
78
79- func: quantized_decomposed::mixed_linear.out(Tensor input, Tensor weight, Tensor weight_scales, Tensor? weight_zero_points, ScalarType? dtype=None, *, Tensor(a!) out) -> Tensor(a!)
80  variants: function
81  kernels:
82    - arg_meta: null
83      kernel_name: torch::executor::quantized_mixed_linear_out
84
85- func: quantized_decomposed::quantize_per_tensor.out(Tensor input, float scale, int zero_point, int quant_min, int quant_max, ScalarType dtype, *, Tensor(a!) out) -> Tensor(a!)
86  variants: function
87  kernels:
88    - arg_meta: null
89      kernel_name: torch::executor::quantize_per_tensor_out
90
91- func: quantized_decomposed::quantize_per_tensor.Tensor_out(Tensor input, Tensor scale, Tensor zero_point, int quant_min, int quant_max, ScalarType dtype, *, Tensor(a!) out) -> Tensor(a!)
92  variants: function
93  kernels:
94    - arg_meta: null
95      kernel_name: torch::executor::quantize_per_tensor_tensor_args_out
96
97- func: quantized_decomposed::choose_qparams_per_token_asymmetric.out(Tensor input, ScalarType dtype, *, Tensor(a!) scale_out, Tensor(b!) zero_point_out) -> (Tensor(a!), Tensor(b!))
98  variants: function
99  kernels:
100    - arg_meta: null
101      kernel_name: torch::executor::choose_qparams_per_token_asymmetric_out
102
103- func: quantized_decomposed::quantize_per_token.out(Tensor input, Tensor scales, Tensor zero_points, int quant_min, int quant_max, ScalarType dtype, *, Tensor(a!) out) -> Tensor(a!)
104  variants: function
105  kernels:
106    - arg_meta: null
107      kernel_name: torch::executor::quantize_per_token_out
108
109- func: quantized_decomposed::dequantize_per_token.out(Tensor input, Tensor scales, Tensor zero_points, int quant_min, int quant_max, ScalarType dtype, ScalarType output_dtype, *, Tensor(a!) out) -> Tensor(a!)
110  variants: function
111  kernels:
112    - arg_meta: null
113      kernel_name: torch::executor::dequantize_per_token_out
114