Skip to content

Commit 5b17e72

Browse files
ggerganovggerganov
committed
docs : add BLAS and Metal ops
1 parent b8a6ff4 commit 5b17e72

File tree

3 files changed

+13154
-86
lines changed

3 files changed

+13154
-86
lines changed

‎docs/ops.md‎

Lines changed: 86 additions & 86 deletions
Original file line numberDiff line numberDiff line change
@@ -7,89 +7,89 @@ Legend:
77
- 🟡 Partially supported by this backend
88
- ❌ Not supported by this backend
99

10-
| Operation | CPU | CUDA |
11-
|-----------|------|------|
12-
| ABS | ✅ | 🟡 |
13-
| ACC | ✅ | ✅ |
14-
| ADD | ✅ | ✅ |
15-
| ADD1 | ✅ | ✅ |
16-
| ARANGE | ✅ | ✅ |
17-
| ARGMAX | ✅ | ✅ |
18-
| ARGSORT | ✅ | ✅ |
19-
| CLAMP | ✅ | ✅ |
20-
| CONCAT | ✅ | 🟡 |
21-
| CONT | ✅ | 🟡 |
22-
| CONV_2D_DW | ✅ | ✅ |
23-
| CONV_TRANSPOSE_1D | ✅ | ✅ |
24-
| CONV_TRANSPOSE_2D | ✅ | ✅ |
25-
| COS | ✅ | ✅ |
26-
| COUNT_EQUAL | ✅ | ✅ |
27-
| CPY | 🟡 | 🟡 |
28-
| CROSS_ENTROPY_LOSS | ✅ | ✅ |
29-
| CROSS_ENTROPY_LOSS_BACK | ✅ | ✅ |
30-
| DIAG_MASK_INF | ✅ | ✅ |
31-
| DIV | ✅ | ✅ |
32-
| DUP | ✅ | 🟡 |
33-
| ELU | ✅ | ❌ |
34-
| EXP | ✅ | 🟡 |
35-
| FLASH_ATTN_EXT | ✅ | 🟡 |
36-
| GATED_LINEAR_ATTN | ✅ | ✅ |
37-
| GEGLU | ✅ | ✅ |
38-
| GEGLU_ERF | ✅ | ✅ |
39-
| GEGLU_QUICK | ✅ | ✅ |
40-
| GELU | ✅ | 🟡 |
41-
| GELU_ERF | ✅ | 🟡 |
42-
| GELU_QUICK | ✅ | 🟡 |
43-
| GET_ROWS | ✅ | 🟡 |
44-
| GET_ROWS_BACK | 🟡 | 🟡 |
45-
| GROUP_NORM | ✅ | ✅ |
46-
| HARDSIGMOID | ✅ | 🟡 |
47-
| HARDSWISH | ✅ | 🟡 |
48-
| IM2COL | ✅ | ✅ |
49-
| L2_NORM | ✅ | ✅ |
50-
| LEAKY_RELU | ✅ | ✅ |
51-
| LOG | ✅ | ✅ |
52-
| MEAN | ✅ | ✅ |
53-
| MUL | ✅ | ✅ |
54-
| MUL_MAT | 🟡 | 🟡 |
55-
| MUL_MAT_ID | ✅ | ✅ |
56-
| NEG | ✅ | 🟡 |
57-
| NORM | ✅ | ✅ |
58-
| OPT_STEP_ADAMW | ✅ | ✅ |
59-
| OUT_PROD | 🟡 | 🟡 |
60-
| PAD | ✅ | ✅ |
61-
| PAD_REFLECT_1D | ✅ | ❌ |
62-
| POOL_2D | ✅ | ✅ |
63-
| REGLU | ✅ | ✅ |
64-
| RELU | ✅ | 🟡 |
65-
| REPEAT | ✅ | 🟡 |
66-
| REPEAT_BACK | ✅ | ✅ |
67-
| RMS_NORM | ✅ | ✅ |
68-
| RMS_NORM_BACK | ✅ | ✅ |
69-
| RMS_NORM_MUL | ✅ | ✅ |
70-
| ROPE | ✅ | ✅ |
71-
| ROPE_BACK | ✅ | ✅ |
72-
| RWKV_WKV6 | ✅ | ✅ |
73-
| RWKV_WKV7 | ✅ | ✅ |
74-
| SCALE | ✅ | ✅ |
75-
| SET | ✅ | ❌ |
76-
| SET_ROWS | 🟡 | ❌ |
77-
| SGN | ✅ | 🟡 |
78-
| SIGMOID | ✅ | 🟡 |
79-
| SILU | ✅ | 🟡 |
80-
| SILU_BACK | ✅ | ✅ |
81-
| SIN | ✅ | ✅ |
82-
| SOFT_MAX | ✅ | ✅ |
83-
| SOFT_MAX_BACK | 🟡 | 🟡 |
84-
| SQR | ✅ | ✅ |
85-
| SQRT | ✅ | ✅ |
86-
| SSM_CONV | ✅ | ✅ |
87-
| SSM_SCAN | ✅ | ✅ |
88-
| STEP | ✅ | 🟡 |
89-
| SUB | ✅ | ✅ |
90-
| SUM | ✅ | ✅ |
91-
| SUM_ROWS | ✅ | ✅ |
92-
| SWIGLU | ✅ | ✅ |
93-
| TANH | ✅ | 🟡 |
94-
| TIMESTEP_EMBEDDING | ✅ | ✅ |
95-
| UPSCALE | ✅ | ✅ |
10+
| Operation | BLAS | CPU | CUDA | Metal |
11+
|-----------|------|------|------|------|
12+
| ABS | ❌ | ✅ | 🟡 | ❌ |
13+
| ACC | ❌ | ✅ | ✅ | ✅ |
14+
| ADD | ❌ | ✅ | ✅ | 🟡 |
15+
| ADD1 | ❌ | ✅ | ✅ | ❌ |
16+
| ARANGE | ❌ | ✅ | ✅ | ✅ |
17+
| ARGMAX | ❌ | ✅ | ✅ | ✅ |
18+
| ARGSORT | ❌ | ✅ | ✅ | ✅ |
19+
| CLAMP | ❌ | ✅ | ✅ | 🟡 |
20+
| CONCAT | ❌ | ✅ | 🟡 | ✅ |
21+
| CONT | ❌ | ✅ | 🟡 | ✅ |
22+
| CONV_2D_DW | ❌ | ✅ | ✅ | ❌ |
23+
| CONV_TRANSPOSE_1D | ❌ | ✅ | ✅ | ✅ |
24+
| CONV_TRANSPOSE_2D | ❌ | ✅ | ✅ | ❌ |
25+
| COS | ❌ | ✅ | ✅ | 🟡 |
26+
| COUNT_EQUAL | ❌ | ✅ | ✅ | ❌ |
27+
| CPY | ❌ | 🟡 | 🟡 | 🟡 |
28+
| CROSS_ENTROPY_LOSS | ❌ | ✅ | ✅ | ❌ |
29+
| CROSS_ENTROPY_LOSS_BACK | ❌ | ✅ | ✅ | ❌ |
30+
| DIAG_MASK_INF | ❌ | ✅ | ✅ | 🟡 |
31+
| DIV | ❌ | ✅ | ✅ | 🟡 |
32+
| DUP | ❌ | ✅ | 🟡 | 🟡 |
33+
| ELU | ❌ | ✅ | ❌ | 🟡 |
34+
| EXP | ❌ | ✅ | 🟡 | ❌ |
35+
| FLASH_ATTN_EXT | ❌ | ✅ | 🟡 | 🟡 |
36+
| GATED_LINEAR_ATTN | ❌ | ✅ | ✅ | ❌ |
37+
| GEGLU | ❌ | ✅ | ✅ | 🟡 |
38+
| GEGLU_ERF | ❌ | ✅ | ✅ | 🟡 |
39+
| GEGLU_QUICK | ❌ | ✅ | ✅ | 🟡 |
40+
| GELU | ❌ | ✅ | 🟡 | 🟡 |
41+
| GELU_ERF | ❌ | ✅ | 🟡 | 🟡 |
42+
| GELU_QUICK | ❌ | ✅ | 🟡 | 🟡 |
43+
| GET_ROWS | ❌ | ✅ | 🟡 | ✅ |
44+
| GET_ROWS_BACK | ❌ | 🟡 | 🟡 | ❌ |
45+
| GROUP_NORM | ❌ | ✅ | ✅ | ✅ |
46+
| HARDSIGMOID | ❌ | ✅ | 🟡 | ❌ |
47+
| HARDSWISH | ❌ | ✅ | 🟡 | ❌ |
48+
| IM2COL | ❌ | ✅ | ✅ | 🟡 |
49+
| L2_NORM | ❌ | ✅ | ✅ | ✅ |
50+
| LEAKY_RELU | ❌ | ✅ | ✅ | ✅ |
51+
| LOG | ❌ | ✅ | ✅ | ❌ |
52+
| MEAN | ❌ | ✅ | ✅ | ✅ |
53+
| MUL | ❌ | ✅ | ✅ | 🟡 |
54+
| MUL_MAT | 🟡 | 🟡 | 🟡 | 🟡 |
55+
| MUL_MAT_ID | ❌ | ✅ | ✅ | ✅ |
56+
| NEG | ❌ | ✅ | 🟡 | 🟡 |
57+
| NORM | ❌ | ✅ | ✅ | 🟡 |
58+
| OPT_STEP_ADAMW | ❌ | ✅ | ✅ | ❌ |
59+
| OUT_PROD | 🟡 | 🟡 | 🟡 | ❌ |
60+
| PAD | ❌ | ✅ | ✅ | ✅ |
61+
| PAD_REFLECT_1D | ❌ | ✅ | ❌ | ✅ |
62+
| POOL_2D | ❌ | ✅ | ✅ | ✅ |
63+
| REGLU | ❌ | ✅ | ✅ | 🟡 |
64+
| RELU | ❌ | ✅ | 🟡 | 🟡 |
65+
| REPEAT | ❌ | ✅ | 🟡 | ✅ |
66+
| REPEAT_BACK | ❌ | ✅ | ✅ | ❌ |
67+
| RMS_NORM | ❌ | ✅ | ✅ | 🟡 |
68+
| RMS_NORM_BACK | ❌ | ✅ | ✅ | ❌ |
69+
| RMS_NORM_MUL | ❌ | ✅ | ✅ | ✅ |
70+
| ROPE | ❌ | ✅ | ✅ | ✅ |
71+
| ROPE_BACK | ❌ | ✅ | ✅ | ❌ |
72+
| RWKV_WKV6 | ❌ | ✅ | ✅ | ✅ |
73+
| RWKV_WKV7 | ❌ | ✅ | ✅ | ✅ |
74+
| SCALE | ❌ | ✅ | ✅ | ✅ |
75+
| SET | ❌ | ✅ | ❌ | ✅ |
76+
| SET_ROWS | ❌ | 🟡 | ❌ | 🟡 |
77+
| SGN | ❌ | ✅ | 🟡 | ❌ |
78+
| SIGMOID | ❌ | ✅ | 🟡 | 🟡 |
79+
| SILU | ❌ | ✅ | 🟡 | 🟡 |
80+
| SILU_BACK | ❌ | ✅ | ✅ | ❌ |
81+
| SIN | ❌ | ✅ | ✅ | 🟡 |
82+
| SOFT_MAX | ❌ | ✅ | ✅ | ✅ |
83+
| SOFT_MAX_BACK | ❌ | 🟡 | 🟡 | ❌ |
84+
| SQR | ❌ | ✅ | ✅ | 🟡 |
85+
| SQRT | ❌ | ✅ | ✅ | 🟡 |
86+
| SSM_CONV | ❌ | ✅ | ✅ | ✅ |
87+
| SSM_SCAN | ❌ | ✅ | ✅ | ✅ |
88+
| STEP | ❌ | ✅ | 🟡 | ❌ |
89+
| SUB | ❌ | ✅ | ✅ | 🟡 |
90+
| SUM | ❌ | ✅ | ✅ | ❌ |
91+
| SUM_ROWS | ❌ | ✅ | ✅ | ✅ |
92+
| SWIGLU | ❌ | ✅ | ✅ | 🟡 |
93+
| TANH | ❌ | ✅ | 🟡 | 🟡 |
94+
| TIMESTEP_EMBEDDING | ❌ | ✅ | ✅ | ✅ |
95+
| UPSCALE | ❌ | ✅ | ✅ | 🟡 |

0 commit comments

Comments
 (0)