Claude/traffic data diffusion model 01 p jfsg4 ss cnw svd hc aqt9ny

IEclaude/check_gpu.py

@@ -0,0 +1,69 @@
+#!/usr/bin/env python3
+"""
+GPU检测脚本
+
+检查系统中可用的GPU设备
+"""
+
+import torch
+
+print("=" * 80)
+print("GPU 设备检测")
+print("=" * 80)
+
+# 检查CUDA是否可用
+if torch.cuda.is_available():
+    print(f"\n✓ CUDA可用")
+    print(f"  CUDA版本: {torch.version.cuda}")
+
+    # 获取GPU数量
+    num_gpus = torch.cuda.device_count()
+    print(f"  可用GPU数量: {num_gpus}")
+
+    # 列出所有GPU
+    print(f"\nGPU详细信息:")
+    for i in range(num_gpus):
+        print(f"\n  GPU {i}:")
+        print(f"    名称: {torch.cuda.get_device_name(i)}")
+
+        # 获取显存信息
+        props = torch.cuda.get_device_properties(i)
+        total_memory = props.total_memory / 1024**3  # 转换为GB
+        print(f"    显存: {total_memory:.2f} GB")
+
+        # 当前显存使用情况
+        if i < num_gpus:
+            torch.cuda.set_device(i)
+            allocated = torch.cuda.memory_allocated(i) / 1024**3
+            reserved = torch.cuda.memory_reserved(i) / 1024**3
+            print(f"    已分配: {allocated:.2f} GB")
+            print(f"    已保留: {reserved:.2f} GB")
+
+    # 使用建议
+    print(f"\n" + "=" * 80)
+    print("使用建议:")
+    print("=" * 80)
+
+    for i in range(num_gpus):
+        gpu_name = torch.cuda.get_device_name(i)
+        props = torch.cuda.get_device_properties(i)
+        total_memory = props.total_memory / 1024**3
+
+        print(f"\nGPU {i} ({gpu_name}):")
+        print(f"  训练命令: python train.py --config configs/config_20.json --gpu {i}")
+        print(f"  或使用脚本: ./run_all.sh --gpu {i}")
+
+        # 根据显存给出batch_size建议
+        if total_memory < 12:
+            print(f"  建议batch_size: 4 (显存较小)")
+        elif total_memory < 24:
+            print(f"  建议batch_size: 8 (默认)")
+        else:
+            print(f"  建议batch_size: 16 (显存充足)")
+
+else:
+    print(f"\n✗ CUDA不可用")
+    print(f"  将使用CPU进行训练（速度会很慢）")
+    print(f"  训练命令: python train.py --config configs/config_20.json")
+
+print("\n" + "=" * 80)

IEclaude/configs/config_20.json

@@ -0,0 +1,61 @@
+{
+    "comment": "IEclaude配置文件 - 20%缺失率",
+
+    "seed": 42,
+
+    "data": {
+        "data_path": "/home/zhu/sssdtcn/LD2011_2014.txt",
+        "seq_len": 168,
+        "stride": 84,
+        "train_ratio": 0.7,
+        "normalize": "standard"
+    },
+
+    "model": {
+        "res_channels": 256,
+        "skip_channels": 256,
+        "num_res_layers": 36,
+        "dilation_cycle": 10,
+
+        "tcn_channels": [256, 256, 256],
+        "tcn_kernel_size": 3,
+        "tcn_dilation_rates": [1, 2, 4, 8],
+        "tcn_dropout": 0.0,
+
+        "s4_d_state": 64,
+        "s4_n_layers": 4,
+        "s4_dropout": 0.0,
+        "s4_bidirectional": true
+    },
+
+    "diffusion": {
+        "T": 200,
+        "beta_0": 0.0001,
+        "beta_T": 0.02,
+        "schedule": "linear",
+
+        "embed_dim_in": 128,
+        "embed_dim_mid": 512,
+        "embed_dim_out": 512
+    },
+
+    "train": {
+        "output_dir": "./results/traffic_20",
+        "missing_rate": 0.2,
+        "missing_pattern": "random",
+
+        "batch_size": 8,
+        "epochs": 100,
+        "learning_rate": 0.0002,
+        "weight_decay": 0.0,
+
+        "scheduler": "cosine",
+        "min_lr": 1e-6,
+
+        "save_interval": 10,
+        "num_workers": 0,
+
+        "only_generate_missing": true,
+        "clean_before_train": true
+    }
+}

IEclaude/data/__init__.py

@@ -0,0 +1,17 @@
+"""
+IEclaude Data Module
+
+包含数据加载和预处理工具
+"""
+
+from .traffic_dataloader import (
+    TrafficDataset,
+    load_traffic_data,
+    create_dataloader
+)
+
+__all__ = [
+    'TrafficDataset',
+    'load_traffic_data',
+    'create_dataloader'
+]