From cab5f98ac60d2573980edc883c0398a876434a17 Mon Sep 17 00:00:00 2001
From: liyang <liyang2@uniontech.com>
Date: Mon, 3 Nov 2025 11:10:03 +0800
Subject: [PATCH 1/2] fix(setup.py): require transformers>=4.53.3 to fix Qwen3
 import

---
 setup.py | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

diff --git a/setup.py b/setup.py
index f83163de..2d466c2a 100644
--- a/setup.py
+++ b/setup.py
@@ -9,18 +9,18 @@
     author_email='yuhui.li@stu.pku.edu.cn',
     url='https://github.com/SafeAILab/EAGLE',
     packages=find_packages(),
-    install_requires=[
-        'torch==2.0.1',
-        'transformers==4.46.2',
-        'accelerate==0.21.0',
-        'fschat==0.2.31',
-        'gradio==3.50.2',
-        'openai==0.28.0',
-        'anthropic==0.5.0',
-        'sentencepiece==0.1.99',
-        'protobuf==3.19.0',
-        'wandb'
-    ],
+    install_requires=[
+        'torch>=2.6.0',
+        'transformers>=4.53.3,<5',
+        'accelerate>=0.26.0',
+        'fschat==0.2.31',
+        'gradio==3.50.2',
+        'openai==0.28.0',
+        'anthropic==0.5.0',
+        'sentencepiece==0.1.99',
+        'protobuf==3.19.0',
+        'wandb'
+    ],
     classifiers=[
         'Development Status :: 3 - Alpha',
         'Intended Audience :: Developers',

From e7238a1d7a50298ed72204473ff772b9e2a7a464 Mon Sep 17 00:00:00 2001
From: liyang <liyang2@uniontech.com>
Date: Mon, 3 Nov 2025 17:10:33 +0800
Subject: [PATCH 2/2] fix(qwen-kv): import GenerationMixin from
 transformers.generation.utils

---
 eagle/model/modeling_qwen2_kv.py | 2 +-
 eagle/model/modeling_qwen3_kv.py | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/eagle/model/modeling_qwen2_kv.py b/eagle/model/modeling_qwen2_kv.py
index 9b8f69de..4f1f91d0 100644
--- a/eagle/model/modeling_qwen2_kv.py
+++ b/eagle/model/modeling_qwen2_kv.py
@@ -29,7 +29,7 @@
 
 from transformers.activations import ACT2FN
 from transformers.cache_utils import Cache, DynamicCache, StaticCache
-from transformers.generation import GenerationMixin
+from transformers.generation.utils import GenerationMixin
 from transformers.modeling_attn_mask_utils import AttentionMaskConverter
 from transformers.modeling_outputs import (
     BaseModelOutputWithPast,
diff --git a/eagle/model/modeling_qwen3_kv.py b/eagle/model/modeling_qwen3_kv.py
index 4a3aeca5..5a22c94c 100644
--- a/eagle/model/modeling_qwen3_kv.py
+++ b/eagle/model/modeling_qwen3_kv.py
@@ -26,7 +26,7 @@
 
 from transformers.activations import ACT2FN
 from transformers.cache_utils import Cache, DynamicCache
-from transformers.generation import GenerationMixin
+from transformers.generation.utils import GenerationMixin
 from transformers.integrations import use_kernel_forward_from_hub
 from transformers.masking_utils import create_causal_mask, create_sliding_window_causal_mask
 from transformers.modeling_flash_attention_utils import FlashAttentionKwargs