Skip to content

ROCm/support GLM-4.7-Flash R3+MTP test #1612

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
sreerohi wants to merge 1 commit into
base: main
Choose a base branch
from
Open

ROCm/support GLM-4.7-Flash R3+MTP test #1612

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions tests/e2e/megatron/test_glm47_flash/_common.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,12 +1,15 @@
import os
from dataclasses import dataclass

import torch

import miles.utils.external_utils.command_utils as U

MODEL_NAME = "GLM-4.7-Flash"
MODEL_TYPE = "glm4.7-flash"

TIGHT_HOST_MEMORY = bool(int(os.environ.get("MILES_TEST_TIGHT_HOST_MEMORY", "1")))
IS_ROCM = torch.version.hip is not None

@gemini-code-assist gemini-code-assist Bot Jul 10, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

Accessing torch.version.hip directly can raise an AttributeError on certain PyTorch builds or older versions where the hip attribute is not defined on the torch.version module. Using getattr is a safer and more robust way to check for ROCm/HIP support.

Suggested change
IS_ROCM = torch.version.hip is not None
IS_ROCM = getattr(torch.version, "hip", None) is not None



@dataclass
Expand Down Expand Up @@ -125,6 +128,8 @@ def build_train_args(case: CaseConfig, *, wandb_file: str) -> str:
"--sglang-speculative-eagle-topk 1 "
"--sglang-speculative-num-draft-tokens 3 "
)
if IS_ROCM:
sglang_args += "--use-miles-router "

if case.use_deepep:
sglang_args += "--sglang-moe-a2a-backend deepep --sglang-deepep-mode auto "
Expand All @@ -134,6 +139,8 @@ def build_train_args(case: CaseConfig, *, wandb_file: str) -> str:
mtp_args = "--enable-mtp-training " "--mtp-loss-scaling-factor 0.2 "

ci_args = "--ci-test "
if IS_ROCM:
ci_args += "--ci-disable-logprobs-checker "

misc_args = (
"--attention-dropout 0.0 "
Expand Down Expand Up @@ -173,8 +180,16 @@ def execute(case: CaseConfig, *, wandb_file: str) -> None:

train_args = build_train_args(case, wandb_file=wandb_file)

extra_env_vars = None
if IS_ROCM:
extra_env_vars = {
"SGLANG_USE_AITER": "0",
"MILES_TEST_R3_THRESHOLD": "1.0",
}

U.execute_train(
train_args=train_args,
num_gpus_per_node=case.num_gpus_per_node,
megatron_model_type=MODEL_TYPE,
extra_env_vars=extra_env_vars,
)