Fix dependency (#538)

python/pyproject.toml

@@ -21,7 +21,7 @@ dependencies = [
 
 [project.optional-dependencies]
 srt = ["aiohttp", "fastapi", "psutil", "rpyc", "torch", "uvloop", "uvicorn",
-       "zmq", "vllm==0.4.3", "interegular", "pydantic", "pillow", "packaging", "huggingface_hub", "hf_transfer", "outlines>=0.0.34"]
+       "zmq", "vllm==0.4.3", "interegular", "pydantic", "pillow", "packaging", "huggingface_hub", "hf_transfer", "outlines>=0.0.41"]
 openai = ["openai>=1.0", "tiktoken"]
 anthropic = ["anthropic>=0.20.0"]
 litellm = ["litellm>=1.0.0"]

python/sglang/srt/constrained/jump_forward.py

@@ -3,10 +3,10 @@ Faster constrained decoding.
 Reference: https://lmsys.org/blog/2024-02-05-compressed-fsm/
 """
 
-import interegular
 import dataclasses
 from collections import defaultdict
 
+import interegular
 import outlines.caching
 from sglang.srt.constrained import (
     FSMInfo,

python/sglang/srt/models/chatglm.py

@@ -5,7 +5,6 @@
 from typing import Iterable, List, Optional, Tuple
 
 import torch
-from peft import LoraConfig
 from sglang.srt.layers.radix_attention import RadixAttention
 from sglang.srt.managers.controller.model_runner import InputMetadata
 from sglang.srt.layers.logits_processor import LogitsProcessor
@@ -31,6 +30,8 @@ from vllm.sequence import SamplerOutput
 from vllm.transformers_utils.configs import ChatGLMConfig
 
 
+LoraConfig = None
+
 
 class GLMAttention(nn.Module):
 
@@ -387,4 +388,4 @@ class ChatGLMForCausalLM(nn.Module):
 
 EntryClass = ChatGLMForCausalLM
 # compat: glm model.config class == ChatGLMModel
-EntryClassRemapping = [("ChatGLMModel", ChatGLMForCausalLM)]
+EntryClassRemapping = [("ChatGLMModel", ChatGLMForCausalLM)]