Skip to content

Commit e9754ab

Browse files
swolchokagrima1304
swolchok
authored and
agrima1304
committed
Bump the PyTorch pin to 20250811 (pytorch#13334)
1 parent 075988d commit e9754ab

File tree

16 files changed

+2132
-2040
lines changed

16 files changed

+2132
-2040
lines changed

‎.ci/docker/ci_commit_pins/pytorch.txt

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1 +1 @@
1-
6fc0ad22f0a07b6f38d138861c56a765d5a9bb02
1+
e7152ff8a6a929a0db7f3f4a72a5b6d471769cd3

‎install_requirements.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -71,7 +71,7 @@ def python_is_compatible():
7171
#
7272
# NOTE: If you're changing, make the corresponding change in .ci/docker/ci_commit_pins/pytorch.txt
7373
# by picking the hash from the same date in https://hud.pytorch.org/hud/pytorch/pytorch/nightly/
74-
NIGHTLY_VERSION = "dev20250725"
74+
NIGHTLY_VERSION = "dev20250811"
7575

7676

7777
def install_requirements(use_pytorch_nightly):

‎runtime/core/portable_type/c10/c10/util/BFloat16-inl.h

Lines changed: 1 addition & 340 deletions
Original file line numberDiff line numberDiff line change
@@ -1,340 +1 @@
1-
#pragma once
2-
3-
#include <c10/macros/Macros.h>
4-
#include <c10/util/bit_cast.h>
5-
6-
#include <limits>
7-
8-
C10_CLANG_DIAGNOSTIC_PUSH()
9-
#if C10_CLANG_HAS_WARNING("-Wimplicit-int-float-conversion")
10-
C10_CLANG_DIAGNOSTIC_IGNORE("-Wimplicit-int-float-conversion")
11-
#endif
12-
13-
#if defined(CL_SYCL_LANGUAGE_VERSION)
14-
#include <CL/sycl.hpp> // for SYCL 1.2.1
15-
#elif defined(SYCL_LANGUAGE_VERSION)
16-
#include <sycl/sycl.hpp> // for SYCL 2020
17-
#endif
18-
19-
namespace c10 {
20-
21-
/// Constructors
22-
inline C10_HOST_DEVICE BFloat16::BFloat16(float value)
23-
:
24-
#if defined(__CUDACC__) && !defined(USE_ROCM) && defined(__CUDA_ARCH__) && \
25-
__CUDA_ARCH__ >= 800
26-
x(__bfloat16_as_ushort(__float2bfloat16(value)))
27-
#elif defined(__SYCL_DEVICE_ONLY__) && \
28-
defined(SYCL_EXT_ONEAPI_BFLOAT16_MATH_FUNCTIONS)
29-
x(c10::bit_cast<uint16_t>(sycl::ext::oneapi::bfloat16(value)))
30-
#else
31-
// RNE by default
32-
x(detail::round_to_nearest_even(value))
33-
#endif
34-
{
35-
}
36-
37-
/// Implicit conversions
38-
inline C10_HOST_DEVICE BFloat16::operator float() const {
39-
#if defined(__CUDACC__) && !defined(USE_ROCM)
40-
return __bfloat162float(*reinterpret_cast<const __nv_bfloat16*>(&x));
41-
#elif defined(__SYCL_DEVICE_ONLY__) && \
42-
defined(SYCL_EXT_ONEAPI_BFLOAT16_MATH_FUNCTIONS)
43-
return float(*reinterpret_cast<const sycl::ext::oneapi::bfloat16*>(&x));
44-
#else
45-
return detail::f32_from_bits(x);
46-
#endif
47-
}
48-
49-
#if defined(__CUDACC__) && !defined(USE_ROCM)
50-
inline C10_HOST_DEVICE BFloat16::BFloat16(const __nv_bfloat16& value) {
51-
x = *reinterpret_cast<const unsigned short*>(&value);
52-
}
53-
inline C10_HOST_DEVICE BFloat16::operator __nv_bfloat16() const {
54-
return *reinterpret_cast<const __nv_bfloat16*>(&x);
55-
}
56-
#endif
57-
58-
#if defined(SYCL_EXT_ONEAPI_BFLOAT16_MATH_FUNCTIONS)
59-
inline C10_HOST_DEVICE BFloat16::BFloat16(
60-
const sycl::ext::oneapi::bfloat16& value) {
61-
x = *reinterpret_cast<const unsigned short*>(&value);
62-
}
63-
inline C10_HOST_DEVICE BFloat16::operator sycl::ext::oneapi::bfloat16() const {
64-
return *reinterpret_cast<const sycl::ext::oneapi::bfloat16*>(&x);
65-
}
66-
#endif
67-
68-
// CUDA intrinsics
69-
70-
#if defined(__CUDACC__) || defined(__HIPCC__)
71-
inline C10_DEVICE BFloat16 __ldg(const BFloat16* ptr) {
72-
#if !defined(USE_ROCM) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 800
73-
return __ldg(reinterpret_cast<const __nv_bfloat16*>(ptr));
74-
#else
75-
return *ptr;
76-
#endif
77-
}
78-
#endif
79-
80-
/// Arithmetic
81-
82-
inline C10_HOST_DEVICE BFloat16
83-
operator+(const BFloat16& a, const BFloat16& b) {
84-
return static_cast<float>(a) + static_cast<float>(b);
85-
}
86-
87-
inline C10_HOST_DEVICE BFloat16
88-
operator-(const BFloat16& a, const BFloat16& b) {
89-
return static_cast<float>(a) - static_cast<float>(b);
90-
}
91-
92-
inline C10_HOST_DEVICE BFloat16
93-
operator*(const BFloat16& a, const BFloat16& b) {
94-
return static_cast<float>(a) * static_cast<float>(b);
95-
}
96-
97-
inline C10_HOST_DEVICE BFloat16 operator/(const BFloat16& a, const BFloat16& b)
98-
__ubsan_ignore_float_divide_by_zero__ {
99-
return static_cast<float>(a) / static_cast<float>(b);
100-
}
101-
102-
inline C10_HOST_DEVICE BFloat16 operator-(const BFloat16& a) {
103-
return -static_cast<float>(a);
104-
}
105-
106-
inline C10_HOST_DEVICE BFloat16& operator+=(BFloat16& a, const BFloat16& b) {
107-
a = a + b;
108-
return a;
109-
}
110-
111-
inline C10_HOST_DEVICE BFloat16& operator-=(BFloat16& a, const BFloat16& b) {
112-
a = a - b;
113-
return a;
114-
}
115-
116-
inline C10_HOST_DEVICE BFloat16& operator*=(BFloat16& a, const BFloat16& b) {
117-
a = a * b;
118-
return a;
119-
}
120-
121-
inline C10_HOST_DEVICE BFloat16& operator/=(BFloat16& a, const BFloat16& b) {
122-
a = a / b;
123-
return a;
124-
}
125-
126-
inline C10_HOST_DEVICE BFloat16& operator|(BFloat16& a, const BFloat16& b) {
127-
a.x = a.x | b.x;
128-
return a;
129-
}
130-
131-
inline C10_HOST_DEVICE BFloat16& operator^(BFloat16& a, const BFloat16& b) {
132-
a.x = a.x ^ b.x;
133-
return a;
134-
}
135-
136-
inline C10_HOST_DEVICE BFloat16& operator&(BFloat16& a, const BFloat16& b) {
137-
a.x = a.x & b.x;
138-
return a;
139-
}
140-
141-
/// Arithmetic with floats
142-
143-
inline C10_HOST_DEVICE float operator+(BFloat16 a, float b) {
144-
return static_cast<float>(a) + b;
145-
}
146-
inline C10_HOST_DEVICE float operator-(BFloat16 a, float b) {
147-
return static_cast<float>(a) - b;
148-
}
149-
inline C10_HOST_DEVICE float operator*(BFloat16 a, float b) {
150-
return static_cast<float>(a) * b;
151-
}
152-
inline C10_HOST_DEVICE float operator/(BFloat16 a, float b) {
153-
return static_cast<float>(a) / b;
154-
}
155-
156-
inline C10_HOST_DEVICE float operator+(float a, BFloat16 b) {
157-
return a + static_cast<float>(b);
158-
}
159-
inline C10_HOST_DEVICE float operator-(float a, BFloat16 b) {
160-
return a - static_cast<float>(b);
161-
}
162-
inline C10_HOST_DEVICE float operator*(float a, BFloat16 b) {
163-
return a * static_cast<float>(b);
164-
}
165-
inline C10_HOST_DEVICE float operator/(float a, BFloat16 b) {
166-
return a / static_cast<float>(b);
167-
}
168-
169-
inline C10_HOST_DEVICE float& operator+=(float& a, const BFloat16& b) {
170-
return a += static_cast<float>(b);
171-
}
172-
inline C10_HOST_DEVICE float& operator-=(float& a, const BFloat16& b) {
173-
return a -= static_cast<float>(b);
174-
}
175-
inline C10_HOST_DEVICE float& operator*=(float& a, const BFloat16& b) {
176-
return a *= static_cast<float>(b);
177-
}
178-
inline C10_HOST_DEVICE float& operator/=(float& a, const BFloat16& b) {
179-
return a /= static_cast<float>(b);
180-
}
181-
182-
/// Arithmetic with doubles
183-
184-
inline C10_HOST_DEVICE double operator+(BFloat16 a, double b) {
185-
return static_cast<double>(a) + b;
186-
}
187-
inline C10_HOST_DEVICE double operator-(BFloat16 a, double b) {
188-
return static_cast<double>(a) - b;
189-
}
190-
inline C10_HOST_DEVICE double operator*(BFloat16 a, double b) {
191-
return static_cast<double>(a) * b;
192-
}
193-
inline C10_HOST_DEVICE double operator/(BFloat16 a, double b) {
194-
return static_cast<double>(a) / b;
195-
}
196-
197-
inline C10_HOST_DEVICE double operator+(double a, BFloat16 b) {
198-
return a + static_cast<double>(b);
199-
}
200-
inline C10_HOST_DEVICE double operator-(double a, BFloat16 b) {
201-
return a - static_cast<double>(b);
202-
}
203-
inline C10_HOST_DEVICE double operator*(double a, BFloat16 b) {
204-
return a * static_cast<double>(b);
205-
}
206-
inline C10_HOST_DEVICE double operator/(double a, BFloat16 b) {
207-
return a / static_cast<double>(b);
208-
}
209-
210-
/// Arithmetic with ints
211-
212-
inline C10_HOST_DEVICE BFloat16 operator+(BFloat16 a, int b) {
213-
return a + static_cast<BFloat16>(b);
214-
}
215-
inline C10_HOST_DEVICE BFloat16 operator-(BFloat16 a, int b) {
216-
return a - static_cast<BFloat16>(b);
217-
}
218-
inline C10_HOST_DEVICE BFloat16 operator*(BFloat16 a, int b) {
219-
return a * static_cast<BFloat16>(b);
220-
}
221-
inline C10_HOST_DEVICE BFloat16 operator/(BFloat16 a, int b) {
222-
return a / static_cast<BFloat16>(b);
223-
}
224-
225-
inline C10_HOST_DEVICE BFloat16 operator+(int a, BFloat16 b) {
226-
return static_cast<BFloat16>(a) + b;
227-
}
228-
inline C10_HOST_DEVICE BFloat16 operator-(int a, BFloat16 b) {
229-
return static_cast<BFloat16>(a) - b;
230-
}
231-
inline C10_HOST_DEVICE BFloat16 operator*(int a, BFloat16 b) {
232-
return static_cast<BFloat16>(a) * b;
233-
}
234-
inline C10_HOST_DEVICE BFloat16 operator/(int a, BFloat16 b) {
235-
return static_cast<BFloat16>(a) / b;
236-
}
237-
238-
//// Arithmetic with int64_t
239-
240-
inline C10_HOST_DEVICE BFloat16 operator+(BFloat16 a, int64_t b) {
241-
return a + static_cast<BFloat16>(b);
242-
}
243-
inline C10_HOST_DEVICE BFloat16 operator-(BFloat16 a, int64_t b) {
244-
return a - static_cast<BFloat16>(b);
245-
}
246-
inline C10_HOST_DEVICE BFloat16 operator*(BFloat16 a, int64_t b) {
247-
return a * static_cast<BFloat16>(b);
248-
}
249-
inline C10_HOST_DEVICE BFloat16 operator/(BFloat16 a, int64_t b) {
250-
return a / static_cast<BFloat16>(b);
251-
}
252-
253-
inline C10_HOST_DEVICE BFloat16 operator+(int64_t a, BFloat16 b) {
254-
return static_cast<BFloat16>(a) + b;
255-
}
256-
inline C10_HOST_DEVICE BFloat16 operator-(int64_t a, BFloat16 b) {
257-
return static_cast<BFloat16>(a) - b;
258-
}
259-
inline C10_HOST_DEVICE BFloat16 operator*(int64_t a, BFloat16 b) {
260-
return static_cast<BFloat16>(a) * b;
261-
}
262-
inline C10_HOST_DEVICE BFloat16 operator/(int64_t a, BFloat16 b) {
263-
return static_cast<BFloat16>(a) / b;
264-
}
265-
266-
// Overloading < and > operators, because std::max and std::min use them.
267-
268-
inline C10_HOST_DEVICE bool operator>(BFloat16& lhs, BFloat16& rhs) {
269-
return float(lhs) > float(rhs);
270-
}
271-
272-
inline C10_HOST_DEVICE bool operator<(BFloat16& lhs, BFloat16& rhs) {
273-
return float(lhs) < float(rhs);
274-
}
275-
276-
} // namespace c10
277-
278-
namespace std {
279-
280-
template <>
281-
class numeric_limits<c10::BFloat16> {
282-
public:
283-
static constexpr bool is_signed = true;
284-
static constexpr bool is_specialized = true;
285-
static constexpr bool is_integer = false;
286-
static constexpr bool is_exact = false;
287-
static constexpr bool has_infinity = true;
288-
static constexpr bool has_quiet_NaN = true;
289-
static constexpr bool has_signaling_NaN = true;
290-
static constexpr auto has_denorm = numeric_limits<float>::has_denorm;
291-
static constexpr auto has_denorm_loss =
292-
numeric_limits<float>::has_denorm_loss;
293-
static constexpr auto round_style = numeric_limits<float>::round_style;
294-
static constexpr bool is_iec559 = false;
295-
static constexpr bool is_bounded = true;
296-
static constexpr bool is_modulo = false;
297-
static constexpr int digits = 8;
298-
static constexpr int digits10 = 2;
299-
static constexpr int max_digits10 = 4;
300-
static constexpr int radix = 2;
301-
static constexpr int min_exponent = -125;
302-
static constexpr int min_exponent10 = -37;
303-
static constexpr int max_exponent = 128;
304-
static constexpr int max_exponent10 = 38;
305-
static constexpr auto traps = numeric_limits<float>::traps;
306-
static constexpr auto tinyness_before =
307-
numeric_limits<float>::tinyness_before;
308-
309-
static constexpr c10::BFloat16 min() {
310-
return c10::BFloat16(0x0080, c10::BFloat16::from_bits());
311-
}
312-
static constexpr c10::BFloat16 lowest() {
313-
return c10::BFloat16(0xFF7F, c10::BFloat16::from_bits());
314-
}
315-
static constexpr c10::BFloat16 max() {
316-
return c10::BFloat16(0x7F7F, c10::BFloat16::from_bits());
317-
}
318-
static constexpr c10::BFloat16 epsilon() {
319-
return c10::BFloat16(0x3C00, c10::BFloat16::from_bits());
320-
}
321-
static constexpr c10::BFloat16 round_error() {
322-
return c10::BFloat16(0x3F00, c10::BFloat16::from_bits());
323-
}
324-
static constexpr c10::BFloat16 infinity() {
325-
return c10::BFloat16(0x7F80, c10::BFloat16::from_bits());
326-
}
327-
static constexpr c10::BFloat16 quiet_NaN() {
328-
return c10::BFloat16(0x7FC0, c10::BFloat16::from_bits());
329-
}
330-
static constexpr c10::BFloat16 signaling_NaN() {
331-
return c10::BFloat16(0x7F80, c10::BFloat16::from_bits());
332-
}
333-
static constexpr c10::BFloat16 denorm_min() {
334-
return c10::BFloat16(0x0001, c10::BFloat16::from_bits());
335-
}
336-
};
337-
338-
} // namespace std
339-
340-
C10_CLANG_DIAGNOSTIC_POP()
1+
#include <torch/headeronly/util/BFloat16.h>

0 commit comments

Comments
 (0)