Skip to content
Permalink

Comparing changes

Choose two branches to see what’s changed or to start a new pull request. If you need to, you can also or learn more about diff comparisons.

Open a pull request

Create a new pull request by comparing changes across two branches. If you need to, you can also . Learn more about diff comparisons here.
base repository: deepspeedai/DeepSpeed
Failed to load repositories. Confirm that selected base ref is valid, then try again.
Loading
base: master
Choose a base ref
...
head repository: abacusai/DeepSpeed
Failed to load repositories. Confirm that selected head ref is valid, then try again.
Loading
compare: master
Choose a head ref
Able to merge. These branches can be automatically merged.
  • 3 commits
  • 1 file changed
  • 1 contributor

Commits on Aug 26, 2024

  1. Allow triton==3.0.x for fp_quantizer 

    Tested with triton==3.0.x and the kernel tests pass so adding as an allowed version.
    @siddartha-RE
    siddartha-RE authored Aug 26, 2024
    Loading
    Loading status checks…

    Verified

    This commit was created on GitHub.com and signed with GitHub’s verified signature.
    GPG key ID: B5690EEEBB952194
    Verified
    Learn about vigilant mode
    Copy the full SHA
    097f525 View commit details

Commits on Aug 27, 2024

  1. Fix variable name

    @siddartha-RE
    siddartha-RE committed Aug 27, 2024
    Loading
    Loading status checks…
    Copy the full SHA
    a628f12 View commit details

  2. Pre-commit fixes

    @siddartha-RE
    siddartha-RE committed Aug 27, 2024
    Loading
    Loading status checks…
    Copy the full SHA
    ed8a765 View commit details
Showing with 10 additions and 6 deletions.
  1. +10 −6 op_builder/fp_quantizer.py
16 changes: 10 additions & 6 deletions op_builder/fp_quantizer.py
View file
Original file line number Diff line number Diff line change
@@ -49,23 +49,27 @@ def is_compatible(self, verbose=False):
import triton
except ImportError:
if verbose:
self.warning(f"please install triton==2.3.0 or 2.3.1 if you want to use the FP Quantizer Kernels")
self.warning(
f"please install triton==2.3.0, 2.3.1 or 3.0.0 if you want to use the FP Quantizer Kernels")
return False

# triton 2.3.0 and 2.3.1 are okay and the only versions released in 2.3.x before 3.x was released
# triton 2.3.{0,1} and 3.0.0 are ok.
allowed_versions = ("2.3", "3.0")
if pkg_version:
allowed = pkg_version.parse("2.3")
allowed = (pkg_version.parse(v) for v in allowed_versions)
installed_triton = pkg_version.parse(triton.__version__)
triton_mismatch = installed_triton.major != allowed.major or installed_triton.minor != allowed.minor
triton_mismatch = all(installed_triton.major != a.major or installed_triton.minor != a.minor
for a in allowed)
else:
installed_triton = triton.__version__
major, minor, _ = installed_triton.split(".")
triton_mismatch = major != "2" or minor != "3"
allowed = (v.split(".") for v in allowed_versions)
triton_mismatch = all(major != v[0] or minor != v[1] for v in allowed)

if triton_mismatch:
if verbose:
self.warning(
f"FP Quantizer is using an untested triton version ({installed_triton}), only 2.3.0 and 2.3.1 are known to be compatible with these kernels"
f"FP Quantizer is using an untested triton version ({installed_triton}), only 2.3.{0,1} and 3.0.0 are known to be compatible with these kernels"
)
return False