Skip to content

Commit 3e2f6bf

Browse files
[NCCL] Upgrade TF NCCL version to 2.25.1
PiperOrigin-RevId: 725521622
1 parent 5dbb1f6 commit 3e2f6bf

File tree

3 files changed

+32
-17
lines changed

3 files changed

+32
-17
lines changed

third_party/gpus/cuda/hermetic/cuda_redist_versions.bzl

+6-6
Original file line numberDiff line numberDiff line change
@@ -133,14 +133,14 @@ CUDNN_REDIST_JSON_DICT = {
133133

134134
CUDA_12_NCCL_WHEEL_DICT = {
135135
"x86_64-unknown-linux-gnu": {
136-
"version": "2.23.4",
137-
"url": "https://files.pythonhosted.org/packages/ed/1f/6482380ec8dcec4894e7503490fc536d846b0d59694acad9cf99f27d0e7d/nvidia_nccl_cu12-2.23.4-py3-none-manylinux2014_x86_64.whl",
138-
"sha256": "b097258d9aab2fa9f686e33c6fe40ae57b27df60cedbd15d139701bb5509e0c1",
136+
"version": "2.25.1",
137+
"url": "https://files.pythonhosted.org/packages/11/0c/8c78b7603f4e685624a3ea944940f1e75f36d71bd6504330511f4a0e1557/nvidia_nccl_cu12-2.25.1-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl",
138+
"sha256": "362aed5963fb9ea2ed2f264409baae30143498fd0e5c503aeaa1badd88cdc54a",
139139
},
140140
"aarch64-unknown-linux-gnu": {
141-
"version": "2.23.4",
142-
"url": "https://files.pythonhosted.org/packages/c8/3a/0112397396dec37ffc8edd7836d48261b4d14ca60ec8ed7bc857cce1d916/nvidia_nccl_cu12-2.23.4-py3-none-manylinux2014_aarch64.whl",
143-
"sha256": "aa946c8327e22ced28e7cef508a334673abc42064ec85f02d005ba1785ea4cec",
141+
"version": "2.25.1",
142+
"url": "https://files.pythonhosted.org/packages/4b/28/f62adab24f2d4b2165b22145af56a7598ab535feb6ccd172f76b9106ebaa/nvidia_nccl_cu12-2.25.1-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl",
143+
"sha256": "4ab428bc915785cc66e8c57cb34c7a64cf739c46702b8db748b6ad6cc7180cf8",
144144
},
145145
}
146146

third_party/nccl/archive.BUILD

+2-2
Original file line numberDiff line numberDiff line change
@@ -22,9 +22,9 @@ exports_files(["LICENSE.txt"])
2222

2323
NCCL_MAJOR = 2
2424

25-
NCCL_MINOR = 23
25+
NCCL_MINOR = 25
2626

27-
NCCL_PATCH = 4
27+
NCCL_PATCH = 1
2828

2929
NCCL_VERSION = NCCL_MAJOR * 10000 + NCCL_MINOR * 100 + NCCL_PATCH # e.g., 21605
3030

third_party/nccl/generated_names.bzl.tpl

+24-9
Original file line numberDiff line numberDiff line change
@@ -6,6 +6,8 @@ GENERATED_SOURCES = [
66
"src/device/generated/all_reduce_minmax_f16.cu.cc",
77
"src/device/generated/all_reduce_minmax_f32.cu.cc",
88
"src/device/generated/all_reduce_minmax_f64.cu.cc",
9+
"src/device/generated/all_reduce_minmax_f8e4m3.cu.cc",
10+
"src/device/generated/all_reduce_minmax_f8e5m2.cu.cc",
911
"src/device/generated/all_reduce_minmax_i32.cu.cc",
1012
"src/device/generated/all_reduce_minmax_i64.cu.cc",
1113
"src/device/generated/all_reduce_minmax_u32.cu.cc",
@@ -15,23 +17,26 @@ GENERATED_SOURCES = [
1517
"src/device/generated/all_reduce_premulsum_f16.cu.cc",
1618
"src/device/generated/all_reduce_premulsum_f32.cu.cc",
1719
"src/device/generated/all_reduce_premulsum_f64.cu.cc",
20+
"src/device/generated/all_reduce_premulsum_f8e4m3.cu.cc",
21+
"src/device/generated/all_reduce_premulsum_f8e5m2.cu.cc",
1822
"src/device/generated/all_reduce_premulsum_u32.cu.cc",
1923
"src/device/generated/all_reduce_premulsum_u64.cu.cc",
2024
"src/device/generated/all_reduce_premulsum_u8.cu.cc",
2125
"src/device/generated/all_reduce_prod_bf16.cu.cc",
2226
"src/device/generated/all_reduce_prod_f16.cu.cc",
2327
"src/device/generated/all_reduce_prod_f32.cu.cc",
2428
"src/device/generated/all_reduce_prod_f64.cu.cc",
29+
"src/device/generated/all_reduce_prod_f8e4m3.cu.cc",
30+
"src/device/generated/all_reduce_prod_f8e5m2.cu.cc",
2531
"src/device/generated/all_reduce_prod_u32.cu.cc",
2632
"src/device/generated/all_reduce_prod_u64.cu.cc",
2733
"src/device/generated/all_reduce_prod_u8.cu.cc",
2834
"src/device/generated/all_reduce_sum_bf16.cu.cc",
2935
"src/device/generated/all_reduce_sum_f16.cu.cc",
3036
"src/device/generated/all_reduce_sum_f32.cu.cc",
3137
"src/device/generated/all_reduce_sum_f64.cu.cc",
32-
"src/device/generated/all_reduce_sumpostdiv_i32.cu.cc",
33-
"src/device/generated/all_reduce_sumpostdiv_i64.cu.cc",
34-
"src/device/generated/all_reduce_sumpostdiv_i8.cu.cc",
38+
"src/device/generated/all_reduce_sum_f8e4m3.cu.cc",
39+
"src/device/generated/all_reduce_sum_f8e5m2.cu.cc",
3540
"src/device/generated/all_reduce_sumpostdiv_u32.cu.cc",
3641
"src/device/generated/all_reduce_sumpostdiv_u64.cu.cc",
3742
"src/device/generated/all_reduce_sumpostdiv_u8.cu.cc",
@@ -46,20 +51,26 @@ GENERATED_SOURCES = [
4651
"src/device/generated/reduce_minmax_f16.cu.cc",
4752
"src/device/generated/reduce_minmax_f32.cu.cc",
4853
"src/device/generated/reduce_minmax_f64.cu.cc",
54+
"src/device/generated/reduce_minmax_f8e4m3.cu.cc",
55+
"src/device/generated/reduce_minmax_f8e5m2.cu.cc",
4956
"src/device/generated/reduce_minmax_u32.cu.cc",
5057
"src/device/generated/reduce_minmax_u64.cu.cc",
5158
"src/device/generated/reduce_minmax_u8.cu.cc",
5259
"src/device/generated/reduce_premulsum_bf16.cu.cc",
5360
"src/device/generated/reduce_premulsum_f16.cu.cc",
5461
"src/device/generated/reduce_premulsum_f32.cu.cc",
5562
"src/device/generated/reduce_premulsum_f64.cu.cc",
63+
"src/device/generated/reduce_premulsum_f8e4m3.cu.cc",
64+
"src/device/generated/reduce_premulsum_f8e5m2.cu.cc",
5665
"src/device/generated/reduce_premulsum_u32.cu.cc",
5766
"src/device/generated/reduce_premulsum_u64.cu.cc",
5867
"src/device/generated/reduce_premulsum_u8.cu.cc",
5968
"src/device/generated/reduce_prod_bf16.cu.cc",
6069
"src/device/generated/reduce_prod_f16.cu.cc",
6170
"src/device/generated/reduce_prod_f32.cu.cc",
6271
"src/device/generated/reduce_prod_f64.cu.cc",
72+
"src/device/generated/reduce_prod_f8e4m3.cu.cc",
73+
"src/device/generated/reduce_prod_f8e5m2.cu.cc",
6374
"src/device/generated/reduce_prod_u32.cu.cc",
6475
"src/device/generated/reduce_prod_u64.cu.cc",
6576
"src/device/generated/reduce_prod_u8.cu.cc",
@@ -68,6 +79,8 @@ GENERATED_SOURCES = [
6879
"src/device/generated/reduce_scatter_minmax_f16.cu.cc",
6980
"src/device/generated/reduce_scatter_minmax_f32.cu.cc",
7081
"src/device/generated/reduce_scatter_minmax_f64.cu.cc",
82+
"src/device/generated/reduce_scatter_minmax_f8e4m3.cu.cc",
83+
"src/device/generated/reduce_scatter_minmax_f8e5m2.cu.cc",
7184
"src/device/generated/reduce_scatter_minmax_i32.cu.cc",
7285
"src/device/generated/reduce_scatter_minmax_i64.cu.cc",
7386
"src/device/generated/reduce_scatter_minmax_u32.cu.cc",
@@ -77,23 +90,26 @@ GENERATED_SOURCES = [
7790
"src/device/generated/reduce_scatter_premulsum_f16.cu.cc",
7891
"src/device/generated/reduce_scatter_premulsum_f32.cu.cc",
7992
"src/device/generated/reduce_scatter_premulsum_f64.cu.cc",
93+
"src/device/generated/reduce_scatter_premulsum_f8e4m3.cu.cc",
94+
"src/device/generated/reduce_scatter_premulsum_f8e5m2.cu.cc",
8095
"src/device/generated/reduce_scatter_premulsum_u32.cu.cc",
8196
"src/device/generated/reduce_scatter_premulsum_u64.cu.cc",
8297
"src/device/generated/reduce_scatter_premulsum_u8.cu.cc",
8398
"src/device/generated/reduce_scatter_prod_bf16.cu.cc",
8499
"src/device/generated/reduce_scatter_prod_f16.cu.cc",
85100
"src/device/generated/reduce_scatter_prod_f32.cu.cc",
86101
"src/device/generated/reduce_scatter_prod_f64.cu.cc",
102+
"src/device/generated/reduce_scatter_prod_f8e4m3.cu.cc",
103+
"src/device/generated/reduce_scatter_prod_f8e5m2.cu.cc",
87104
"src/device/generated/reduce_scatter_prod_u32.cu.cc",
88105
"src/device/generated/reduce_scatter_prod_u64.cu.cc",
89106
"src/device/generated/reduce_scatter_prod_u8.cu.cc",
90107
"src/device/generated/reduce_scatter_sum_bf16.cu.cc",
91108
"src/device/generated/reduce_scatter_sum_f16.cu.cc",
92109
"src/device/generated/reduce_scatter_sum_f32.cu.cc",
93110
"src/device/generated/reduce_scatter_sum_f64.cu.cc",
94-
"src/device/generated/reduce_scatter_sumpostdiv_i32.cu.cc",
95-
"src/device/generated/reduce_scatter_sumpostdiv_i64.cu.cc",
96-
"src/device/generated/reduce_scatter_sumpostdiv_i8.cu.cc",
111+
"src/device/generated/reduce_scatter_sum_f8e4m3.cu.cc",
112+
"src/device/generated/reduce_scatter_sum_f8e5m2.cu.cc",
97113
"src/device/generated/reduce_scatter_sumpostdiv_u32.cu.cc",
98114
"src/device/generated/reduce_scatter_sumpostdiv_u64.cu.cc",
99115
"src/device/generated/reduce_scatter_sumpostdiv_u8.cu.cc",
@@ -104,9 +120,8 @@ GENERATED_SOURCES = [
104120
"src/device/generated/reduce_sum_f16.cu.cc",
105121
"src/device/generated/reduce_sum_f32.cu.cc",
106122
"src/device/generated/reduce_sum_f64.cu.cc",
107-
"src/device/generated/reduce_sumpostdiv_i32.cu.cc",
108-
"src/device/generated/reduce_sumpostdiv_i64.cu.cc",
109-
"src/device/generated/reduce_sumpostdiv_i8.cu.cc",
123+
"src/device/generated/reduce_sum_f8e4m3.cu.cc",
124+
"src/device/generated/reduce_sum_f8e5m2.cu.cc",
110125
"src/device/generated/reduce_sumpostdiv_u32.cu.cc",
111126
"src/device/generated/reduce_sumpostdiv_u64.cu.cc",
112127
"src/device/generated/reduce_sumpostdiv_u8.cu.cc",

0 commit comments

Comments
 (0)