SEO-Clustering-Tool/requirements-semantic-amd.txt at main · FassihFayyaz/SEO-Clustering-Tool · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
# Semantic Clustering Dependencies - AMD GPU (Linux Only)
#
# This file installs semantic clustering dependencies for AMD GPUs
# Provides 3-7x performance improvement over CPU-only processing
#
# IMPORTANT: Install PyTorch with ROCm FIRST, then this file!
#
# Step 1: Install PyTorch with ROCm support (Linux only)
# pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.6
#
# Step 2: Install this requirements file
# pip install -r requirements-semantic-amd.txt
#
# Hardware Requirements:
# - GPU: AMD GPU with ROCm support (RX 6000/7000 series, MI series)
# - OS: Linux only (ROCm not supported on Windows)
# - VRAM: 4GB+ (6GB+ recommended)
# - RAM: 4GB+ system RAM
# - Storage: ~1-2GB for dependencies + 2-10GB for models
#
# Performance:
# - Speed: 3-7x faster than CPU
# - Models: All models supported with appropriate VRAM
# - Compatibility: AMD GPUs with ROCm support on Linux

# =============================================================================
# Important: Linux Only!
# =============================================================================
#
# ROCm (AMD's GPU computing platform) is only supported on Linux
# If you're on Windows, please use requirements-semantic-cpu.txt instead
#
# Supported AMD GPUs:
# - RX 6000 series (RX 6600, 6700, 6800, 6900)
# - RX 7000 series (RX 7600, 7700, 7800, 7900)
# - MI series (MI100, MI200, MI300)
# - Some older cards with ROCm support

# =============================================================================
# Semantic Clustering Dependencies (PyTorch with ROCm must be installed first!)
# =============================================================================

# Main semantic clustering library
sentence-transformers>=2.2.0

# Transformer models and tokenizers
transformers>=4.30.0

# Model acceleration and optimization
accelerate>=0.20.0

# Safe tensor format for faster model loading
safetensors>=0.3.0

# =============================================================================
# Optional Performance Enhancements
# =============================================================================

# Optimized model inference (uncomment if needed)
# optimum>=1.12.0

# =============================================================================
# Installation Notes
# =============================================================================

# Total Download Size: ~1-2GB (includes ROCm libraries)
#
# VRAM Requirements by Model:
# ✅ gte-large: 4GB+ VRAM
# ✅ bge-large-en-v1.5: 4GB+ VRAM
# ✅ sentence-t5-xl: 8GB+ VRAM
# ✅ Qwen-0.6B: 6GB+ VRAM
# ✅ Qwen-4B: 12GB+ VRAM
# ✅ Qwen-8B: 24GB+ VRAM
#
# Performance Tips:
# - Monitor GPU memory with: rocm-smi
# - AMD GPUs generally need more VRAM than NVIDIA for same models
# - Use gte-large or bge-large for best compatibility
# - Consider CPU fallback if VRAM is insufficient
#
# Troubleshooting:
# - If ROCm not detected: check AMD GPU drivers and ROCm installation
# - If out of memory: use smaller model or CPU version
# - If installation fails: ensure you're on Linux with supported AMD GPU
# - For Windows users: use requirements-semantic-cpu.txt instead
#
# ROCm Installation (if needed):
# Follow AMD's official ROCm installation guide for your Linux distribution
# https://rocm.docs.amd.com/en/latest/deploy/linux/index.html