[Chore] Use upstream Transformers DeepseekV4Config (#61392)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
Co-authored-by: Kevin H. Luu <khluu000@gmail.com>
This commit is contained in:
Harry Mellor
2026-10-15 01:08:08 -07:00
committed by GitHub
co-authored by Claude Opus 5.5 mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> Kevin H. Luu
parent 1c4c319792
commit b965ee564d
31 changed files with 155 additions and 120 deletions
+2 -4
View File
@@ -2005,9 +2005,7 @@ def test_draft_inherits_ep_only_for_moe(
draft_is_moe: bool,
):
"""Validate final draft configs without loading weights or mocking validation."""
from transformers import LlamaConfig, MixtralConfig
from vllm.transformers_utils.configs.deepseek_v4 import DeepseekV4Config
from transformers import DeepseekV4Config, LlamaConfig, MixtralConfig
common = dict(
hidden_size=128,
@@ -2024,7 +2022,7 @@ def test_draft_inherits_ep_only_for_moe(
n_routed_experts=4,
num_experts_per_tok=2,
num_nextn_predict_layers=1,
compress_ratios=[1, 1],
layer_types=["sliding_attention"] * 2,
head_dim=32,
**common,
)