weekii
ffc3cdd339
Added test suite to verify Ollama and Nexa SDK API integration: Test files: 1. tests/test_ollama_nexa_comparison.py - Compares Ollama vs Nexa SDK APIs - Tests service availability - Tests model listing - Tests text generation 2. tests/test_unified_text_node.py - Tests unified node integration - Tests all three modes: Local GGUF, Ollama, Nexa SDK - Verifies parameter passing Test results (2025-10-29): ✅ Ollama API (127.0.0.1:11435) - Model: huihui-qwen3-4b:latest - Prompt: 用中文说你好 - Output: 你好！ ✅ Nexa SDK API (127.0.0.1:11434) - Model: mradermacher/Huihui-Qwen3-4B-Instruct-2507-abliterated-GGUF:Q8_0 - Prompt: 用中文说你好 - Output: 你好！ Key findings: - Both APIs use identical OpenAI-compatible format - Same GGUF model works in both backends - Unified API engine successfully handles both - Only difference is endpoint URL and model naming Ollama setup for testing: ```bash # Install Ollama curl -fsSL https://ollama.com/install.sh | sh # Create Modelfile cat > Modelfile << EOF FROM /path/to/model.gguf TEMPLATE """...""
131 lines
3.2 KiB
Python
131 lines
3.2 KiB
Python
#!/usr/bin/env python3
|
|
"""
|
|
测试统一文本生成节点
|
|
测试 Local GGUF、Ollama API、Nexa SDK API
|
|
"""
|
|
|
|
import sys
|
|
import os
|
|
|
|
# 添加 ComfyUI 路径
|
|
sys.path.insert(0, '/workspace/ComfyUI')
|
|
sys.path.insert(0, '/workspace/ComfyUI/custom_nodes/ComfyUI-GGUF-FX')
|
|
|
|
# 模拟 folder_paths
|
|
class FolderPaths:
|
|
models_dir = "/workspace/ComfyUI/models"
|
|
|
|
sys.modules['folder_paths'] = FolderPaths()
|
|
|
|
from nodes.unified_text_node import UnifiedTextModelSelector, UnifiedTextGeneration
|
|
|
|
print("="*80)
|
|
print(" 测试统一文本生成节点")
|
|
print("="*80)
|
|
|
|
# 测试 1: Ollama API
|
|
print("\n" + "="*80)
|
|
print("测试 1: Ollama API")
|
|
print("="*80)
|
|
|
|
selector = UnifiedTextModelSelector()
|
|
config, = selector.select_model(
|
|
mode="Remote (API)",
|
|
base_url="http://127.0.0.1:11435",
|
|
api_type="Ollama",
|
|
remote_model="",
|
|
refresh_models=False,
|
|
system_prompt="You are a helpful assistant."
|
|
)
|
|
|
|
if config.get("service_available"):
|
|
generator = UnifiedTextGeneration()
|
|
context, thinking = generator.generate(
|
|
model_config=config,
|
|
prompt="用中文介绍一下你自己，不超过50字",
|
|
max_tokens=100,
|
|
temperature=0.7,
|
|
top_p=0.9,
|
|
top_k=40,
|
|
repetition_penalty=1.1,
|
|
enable_thinking=False
|
|
)
|
|
|
|
print(f"\n📝 生成结果:")
|
|
print(f" {context}")
|
|
else:
|
|
print("❌ Ollama 服务不可用")
|
|
|
|
# 测试 2: Nexa SDK API
|
|
print("\n" + "="*80)
|
|
print("测试 2: Nexa SDK API")
|
|
print("="*80)
|
|
|
|
config, = selector.select_model(
|
|
mode="Remote (API)",
|
|
base_url="http://127.0.0.1:11434",
|
|
api_type="Nexa SDK",
|
|
remote_model="",
|
|
refresh_models=False,
|
|
system_prompt="You are a helpful assistant."
|
|
)
|
|
|
|
if config.get("service_available"):
|
|
generator = UnifiedTextGeneration()
|
|
context, thinking = generator.generate(
|
|
model_config=config,
|
|
prompt="Say hello in Chinese",
|
|
max_tokens=50,
|
|
temperature=0.7,
|
|
top_p=0.9,
|
|
top_k=40,
|
|
repetition_penalty=1.1,
|
|
enable_thinking=False
|
|
)
|
|
|
|
print(f"\n📝 生成结果:")
|
|
print(f" {context}")
|
|
else:
|
|
print("❌ Nexa SDK 服务不可用")
|
|
|
|
# 测试 3: Local GGUF
|
|
print("\n" + "="*80)
|
|
print("测试 3: Local GGUF")
|
|
print("="*80)
|
|
|
|
gguf_file = "Huihui-Qwen3-4B-Instruct-2507-abliterated.Q8_0.gguf"
|
|
gguf_path = f"/workspace/ComfyUI/models/LLM/GGUF/{gguf_file}"
|
|
|
|
if os.path.exists(gguf_path):
|
|
config, = selector.select_model(
|
|
mode="Local (GGUF)",
|
|
local_model=gguf_file,
|
|
n_ctx=8192,
|
|
n_gpu_layers=-1,
|
|
system_prompt="You are a helpful assistant."
|
|
)
|
|
|
|
if "error" not in config:
|
|
generator = UnifiedTextGeneration()
|
|
context, thinking = generator.generate(
|
|
model_config=config,
|
|
prompt="用中文说你好",
|
|
max_tokens=50,
|
|
temperature=0.7,
|
|
top_p=0.9,
|
|
top_k=40,
|
|
repetition_penalty=1.1,
|
|
enable_thinking=False
|
|
)
|
|
|
|
print(f"\n📝 生成结果:")
|
|
print(f" {context}")
|
|
else:
|
|
print(f"❌ {config['error']}")
|
|
else:
|
|
print(f"❌ GGUF 文件不存在: {gguf_path}")
|
|
|
|
print("\n" + "="*80)
|
|
print(" 测试完成!")
|
|
print("="*80)
|