DeepGPR 0.0.2__tar.gz → 0.0.5__tar.gz

{deepgpr-0.0.2/DeepGPR.egg-info → deepgpr-0.0.5}/PKG-INFO

@@ -1,10 +1,11 @@
 Metadata-Version: 2.4
 Name: DeepGPR
-Version: 0.0.2
+Version: 0.0.5
 Summary: PyTorch and CUDA for GPR FWI
 Author-email: Lei Liu <liulei990222@gmail.com>
 Classifier: Programming Language :: Python :: 3
-Classifier: Operating System :: OS Independent
+Classifier: Operating System :: Microsoft :: Windows
+Classifier: Operating System :: POSIX :: Linux
 Requires-Python: >=3.7
 Description-Content-Type: text/markdown
 
@@ -59,7 +60,6 @@ peak_time = 1 / freq
 source_amplitudes = torch.zeros((1,nt,1),device=device)
 source_amplitudes[0,:,0]=DeepGPR.ricker(freq, nt, dt, peak_time).to(device)
 
-DeepGPR.plot_survey_geometry(er,source_location, receiver_location,dx)
 
 #forward modeling
 r = DeepGPR.compute(
@@ -79,8 +79,7 @@ ax[1].imshow(er.grad.detach())
 ax[1].set_title("Gradient")
 plt.show()
 ```
-![result](./Fig/example1.png)
 
-![result](./Fig/example2.png)
+![result](./Fig/example.png)
 
 There are more examples in the ./examples.

{deepgpr-0.0.2 → deepgpr-0.0.5}/README.md

@@ -49,7 +49,6 @@ peak_time = 1 / freq
 source_amplitudes = torch.zeros((1,nt,1),device=device)
 source_amplitudes[0,:,0]=DeepGPR.ricker(freq, nt, dt, peak_time).to(device)
 
-DeepGPR.plot_survey_geometry(er,source_location, receiver_location,dx)
 
 #forward modeling
 r = DeepGPR.compute(
@@ -69,8 +68,7 @@ ax[1].imshow(er.grad.detach())
 ax[1].set_title("Gradient")
 plt.show()
 ```
-![result](./Fig/example1.png)
 
-![result](./Fig/example2.png)
+![result](./Fig/example.png)
 
 There are more examples in the ./examples.

{deepgpr-0.0.2 → deepgpr-0.0.5}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "DeepGPR" 
-version = "0.0.2"              
+version = "0.0.5"              
 authors = [
   { name="Lei Liu", email="liulei990222@gmail.com" },
 ]
@@ -13,11 +13,13 @@ readme = "README.md"
 requires-python = ">=3.7"
 classifiers = [
     "Programming Language :: Python :: 3",
-    "Operating System :: OS Independent"
+    "Operating System :: Microsoft :: Windows",
+    "Operating System :: POSIX :: Linux"
 ]
 
-[tool.setuptools]
-packages = ["DeepGPR"]
+
+[tool.setuptools.packages.find]
+where = ["src"]
 
 [tool.setuptools.package-data]
 "DeepGPR" = ["lib/*.cu", "lib/*.so", "lib/*.dll"]

{deepgpr-0.0.2 → deepgpr-0.0.5/src}/DeepGPR/__init__.py

@@ -14,7 +14,6 @@ if system_name == "Windows":
     lib_filename = f'deepgpr{lib_extension}'
     lib_path_obj = lib_dir / lib_filename
     
-    # Windows 下直接调用 nvcc 的命令参数
     nvcc_cmd = [
         'nvcc', '-shared', 
         '-o', str(lib_path_obj), 
@@ -25,7 +24,6 @@ else:
     lib_filename = f'deepgpr{lib_extension}'
     lib_path_obj = lib_dir / lib_filename
     
-    # Linux 下直接调用 nvcc 的命令参数 (-fPIC 和 ABI 设置)
     nvcc_cmd = [
         'nvcc', '-shared', '-Xcompiler', '-fPIC', 
         '-D_GLIBCXX_USE_CXX11_ABI=0', 
@@ -33,11 +31,9 @@ else:
         str(cu_file)
     ]
 
-# 2. 检查动态库是否存在，不存在则直接使用 Python 调用 nvcc 进行编译
 if not lib_path_obj.is_file():
     print(f'Compiling CUDA extension for {system_name} directly via nvcc...')
     try:
-        # 直接执行 nvcc 命令，完全摆脱 make 依赖
         subprocess.run(nvcc_cmd, check=True)
     except FileNotFoundError:
         raise RuntimeError(
@@ -47,11 +43,10 @@ if not lib_path_obj.is_file():
     except subprocess.CalledProcessError as e:
         raise RuntimeError(f"Compilation failed with error code {e.returncode}.")
 
-# 3. 加载编译好的动态链接库 (变量名改为 c_lib，避免与 lib 文件夹冲突)
 lib_path = str(lib_path_obj)
 c_lib = ctypes.cdll.LoadLibrary(lib_path)
 
-# 4. 定义 C 函数的参数和返回值类型
+
 c_lib.forward.argtypes = [
     ctypes.POINTER(ctypes.c_float), ctypes.POINTER(ctypes.c_float), 
     ctypes.POINTER(ctypes.c_float), ctypes.POINTER(ctypes.c_float), 
@@ -92,7 +87,7 @@ c_lib.forward.argtypes = [
     ctypes.POINTER(ctypes.c_int), ctypes.POINTER(ctypes.c_float),
     ctypes.c_int 
 ]
-c_lib.forward.restype = None  # 修复了绑定错误
+c_lib.forward.restype = None  
 
 c_lib.backward.argtypes = [
     ctypes.POINTER(ctypes.c_float), ctypes.POINTER(ctypes.c_float), 
@@ -134,14 +129,11 @@ c_lib.backward.argtypes = [
     ctypes.c_int ,ctypes.POINTER(ctypes.c_float), ctypes.POINTER(ctypes.c_float),
     ctypes.c_int, ctypes.c_int 
 ]
-c_lib.backward.restype = None  # 修复了绑定错误
-
+c_lib.backward.restype = None 
 __all__ = ['c_lib']
 
-# ==============================================================================
-# 第二步：在动态库环境就绪后再导入其余模块
-# ==============================================================================
+
 from .common import *
 from .compute2 import *
 from .multiscale import *
-from .visual import *
+from .wavelet import *

{deepgpr-0.0.2 → deepgpr-0.0.5/src}/DeepGPR/common.py

@@ -5,37 +5,7 @@ import math
 from scipy.constants import c
 from scipy.constants import mu_0 as m0
 from scipy.constants import epsilon_0 as e0
-from typing import Optional
-
-def ricker(
-    freq: float,
-    length: int,
-    dt: float,
-    peak_time: float,
-    dtype: Optional[torch.dtype] = None,
-) -> torch.Tensor:
-    """Return a Ricker wavelet with the specified central frequency.
-
-    Args:
-        freq: The central frequency.
-        length: The number of time samples.
-        dt: The time sample spacing.
-        peak_time: The time (in secs) of the peak amplitude.
-        dtype: The PyTorch datatype to use. Optional, defaults to PyTorch's
-            default (float32).
-    Returns:
-        A PyTorch tensor representing the Ricker wavelet.
-    """
-    if dt == 0:
-        raise ValueError("dt cannot be zero.")
 
-    t: torch.Tensor = torch.arange(float(length), dtype=dtype) * dt - peak_time
-    y: torch.Tensor = (1 - 2 * math.pi**2 * freq**2 * t**2) * torch.exp(
-        -(math.pi**2) * freq**2 * t**2,
-    )
-    if dtype is not None:
-        return y.to(dtype)
-    return y
 
 def initialization(device, er,se,mr,source_amplitudes,source_location,receiver_location,dx,dt,pmlthick):
     dtype=torch.float32
@@ -150,7 +120,7 @@ def check_cfl(dx, dt, nx,ny,nz):
 
 
 def pmlthick_revert(p, er):
-    if isinstance(p, int):  # 如果是 int
+    if isinstance(p, int):  
         if er.shape[2] == 1:
             return torch.tensor([p, p, p, p, 0, 0], dtype=torch.int32)
         return torch.tensor([p]*6, dtype=torch.int32)
@@ -310,7 +280,7 @@ def check_tensors_for_nan_inf(d,**tensors):
 
         if has_nan or has_inf:
             found_issue = True
-            print(f"❌ [ERROR]{d}: Tensor `{name}` contains:", end=" ")
+            print(f"[ERROR]{d}: Tensor `{name}` contains:", end=" ")
             if has_nan:
                 print("NaN ", end="")
             if has_inf:
@@ -594,50 +564,3 @@ def checkpoint_initial_field(device=None,per_nstep=None, dx=None, dt=None,
 
 def zero_field(*tensors):
     return tuple(torch.zeros_like(t) if t is not None else None for t in tensors)
-
-def print_field_shapes(E, H, PML):
-    """
-    打印 FDTD 场变量的形状
-    E: (Ex, Ey, Ez)
-    H: (Hx, Hy, Hz)
-    PML: 包含 24 个 Phi 张量的元组
-    """
-    print("="*30)
-    print(" FIELD SHAPES CHECK ")
-    print("="*30)
-    
-    # 1. 检查电场 E
-    e_names = ['Ex', 'Ey', 'Ez']
-    for name, tensor in zip(e_names, E):
-        shape = tensor.shape if torch.is_tensor(tensor) else "Not a Tensor"
-        print(f"{name:10} : {shape}")
-    
-    print("-" * 20)
-    
-    # 2. 检查磁场 H
-    h_names = ['Hx', 'Hy', 'Hz']
-    for name, tensor in zip(h_names, H):
-        shape = tensor.shape if torch.is_tensor(tensor) else "Not a Tensor"
-        print(f"{name:10} : {shape}")
-    
-    print("-" * 20)
-    
-    # 3. 检查 PML 场
-    pml_names = [
-        "x0EPhi1", "x0EPhi2", "x0HPhi1", "x0HPhi2",
-        "xmEPhi1", "xmEPhi2", "xmHPhi1", "xmHPhi2",
-        "y0EPhi1", "y0EPhi2", "y0HPhi1", "y0HPhi2",
-        "ymEPhi1", "ymEPhi2", "ymHPhi1", "ymHPhi2",
-        "z0EPhi1", "z0EPhi2", "z0HPhi1", "z0HPhi2",
-        "zmEPhi1", "zmEPhi2", "zmHPhi1", "zmHPhi2"
-    ]
-    
-    for i, name in enumerate(pml_names):
-        if i < len(PML):
-            tensor = PML[i]
-            shape = tensor.shape if torch.is_tensor(tensor) else "Not a Tensor"
-            print(f"{name:10} : {shape}")
-        else:
-            print(f"{name:10} : Missing in PML tuple")
-            
-    print("="*30)

{deepgpr-0.0.2 → deepgpr-0.0.5/src}/DeepGPR/compute2.py

@@ -60,7 +60,6 @@ class DeepGPR(torch.autograd.Function):
 
         nt_saved = (nt + model_gradient_sampling_interval - 1) // model_gradient_sampling_interval
         
-        # 仅仅控制显存分配策略，无需修改 C 端参数签名
         if use_async_offload:
             Eall = torch.zeros((nt_saved, nstep, nx, ny, nz), device='cpu', dtype=dtype).pin_memory()
         else:
@@ -307,7 +306,6 @@ class DeepGPR(torch.autograd.Function):
         ctx.Eall = None
         del Eall,er, se, mr,receiver_location,x0,xm,y0,ym,z0,zm,x01,x02,xm1,xm2,y01,y02,ym1,ym2,z01,z02,zm1,zm2,ere,see, Eupdatecoffs0, Eupdatecoffs1, Eupdatecoffs4, Hupdatecoffs0, Hupdatecoffs1, Hupdatecoffs4
 
-        # 返回与前向传播参数一一对应（最后补齐 use_async_offload 占位）
         return (
                     grad_er, grad_se,         
                     gEx,gEy,gEz, gHx,gHy,gHz, 

{deepgpr-0.0.2 → deepgpr-0.0.5/src}/DeepGPR/lib/deepgpr.cu

@@ -17,9 +17,6 @@ __constant__ float m0 = 1.25663706212e-06;
     }\
 }
 
-// ---------------------------------------------------------
-// 系数获取核函数
-// ---------------------------------------------------------
 __global__ void ucgetforward(const float* __restrict__ er, const float* __restrict__ se, const float* __restrict__ mr,
     float* __restrict__ uE0, float* __restrict__ uE1, float* __restrict__ uE4,
     float* __restrict__ uH0, float* __restrict__ uH1, float* __restrict__ uH4,
@@ -85,9 +82,7 @@ __global__ void ucgetbackward(const float* __restrict__ er, const float* __restr
     }
 }
 
-// ---------------------------------------------------------
-// 接收端保存
-// ---------------------------------------------------------
+
 __global__ void store_outputs(
     int step, int NRX, int iteration,
     const int* __restrict__ receiverlocation, float* __restrict__ rxs,
@@ -116,9 +111,7 @@ __global__ void store_outputs(
     }
 }
 
-// ---------------------------------------------------------
-// 震源更新
-// ---------------------------------------------------------
+
 __global__ void Update_hertzian_dipole(
     int step, int iteration, float dx, 
     const int* __restrict__ sourcelocation, const float* __restrict__ srcwaveforms,
@@ -146,9 +139,7 @@ __global__ void Update_hertzian_dipole(
     }
 }
 
-// ---------------------------------------------------------
-// 融合：电场全局更新 + 全侧 PML 边界修正
-// ---------------------------------------------------------
+
 __global__ void fused_e_fields_updates_gpu(
     const float* __restrict__ uE0, const float* __restrict__ uE1,  
     float* __restrict__ Ex, float* __restrict__ Ey, float* __restrict__ Ez,  
@@ -317,9 +308,7 @@ __global__ void fused_e_fields_updates_gpu(
     }
 }
 
-// ---------------------------------------------------------
-// 融合：磁场全局更新 + 全侧 PML 边界修正
-// ---------------------------------------------------------
+
 __global__ void fused_h_fields_updates_gpu(
     const float* __restrict__ uH0, const float* __restrict__ uH1,
     const float* __restrict__ Ex, const float* __restrict__ Ey, const float* __restrict__ Ez,
@@ -488,9 +477,7 @@ __global__ void fused_h_fields_updates_gpu(
     }
 }
 
-// ---------------------------------------------------------
-// 反传：波场倒播
-// ---------------------------------------------------------
+
 __global__ void Back_source(
     int step, int iteration, float dx,
     const int* __restrict__ sourcelocation, const float* __restrict__ srcwaveforms,
@@ -519,9 +506,7 @@ __global__ void Back_source(
     }
 }
 
-// ---------------------------------------------------------
-// 提取快照到缓冲区或全局内存
-// ---------------------------------------------------------
+
 __global__ void copy_to_Eall_single(
     float* __restrict__ dst_ptr, int t_idx, const float* __restrict__ E, 
     int step, int NX, int NY, int NZ)
@@ -547,9 +532,7 @@ __global__ void copy_to_Eall_single(
     }
 }
 
-// ---------------------------------------------------------
-// 融合：伴随状态法梯度更新（支持降采样波场及异步滑窗/同步显存自适应）
-// ---------------------------------------------------------
+
 __global__ void accumulate_gradients(
     const float* __restrict__ Ez, const float* __restrict__ Eall_ptr, const float* __restrict__ d_E_buf,
     float* __restrict__ grader, float* __restrict__ gradse,
@@ -569,7 +552,6 @@ __global__ void accumulate_gradients(
 
     long long idx_Ez = ix * NY * NZ + iy * NZ + iz;
     
-    // 逻辑时刻索引
     long long idx0_curr = i / S;
     long long idx1_curr = min(idx0_curr + 1, (long long)nt_saved - 1);
     float w1_curr = (float)(i % S) / S;
@@ -615,9 +597,7 @@ __global__ void accumulate_gradients(
     if (serequiregrad == 1) atomicAdd(&gradse[idx], local_gradse);
 }
 
-// ---------------------------------------------------------
-// 主机 API
-// ---------------------------------------------------------
+
 extern "C" {
 
 void forward(const float* __restrict__ er, const float* __restrict__ se, const float* __restrict__ mr,  
@@ -713,7 +693,6 @@ void forward(const float* __restrict__ er, const float* __restrict__ se, const f
     }
 
     if (use_async) {
-        // 【核心修复】：必须先等流里的所有任务彻底做完，才能把底下的显存 free 掉！
         cudaStreamSynchronize(stream_comp);
         cudaStreamSynchronize(stream_trans);
         cudaFree(d_E_buf); 
@@ -829,7 +808,6 @@ void backward(const float* __restrict__ er, const float* __restrict__ se, const
     }
 
     if (use_async) {
-        // 【核心修复】：必须先同步！
         cudaStreamSynchronize(stream_comp);
         cudaStreamSynchronize(stream_trans);
         cudaFree(d_E_buf); 

{deepgpr-0.0.2 → deepgpr-0.0.5/src}/DeepGPR/multiscale.py

@@ -2,28 +2,21 @@
 import torch
 
 def design_fir_filter(cutoff: float, fs: float, numtaps: int) -> torch.Tensor:
-    """使用 Hamming 窗设计 FIR 滤波器"""
     n = torch.arange(numtaps, dtype=torch.float32)
-    # 汉明窗
     window = 0.54 - 0.46 * torch.cos(2 * torch.pi * n / (numtaps - 1))
-    # 正弦波
     sinc = torch.sin(2 * torch.pi * (cutoff/fs) * (n - (numtaps-1)/2)) / (torch.pi * (n - (numtaps-1)/2))
-    # 处理中心点
     center = (numtaps-1) // 2
     sinc[center] = 2 * cutoff/fs
-    # 应用窗函数
     h = window * sinc
-    # 归一化
+
     return h / h.sum()
 
 def apply_filter(data: torch.Tensor, fs: float, cutoff: float) -> torch.Tensor:
-    """应用 FIR 滤波器到数据"""
     numtaps = int(1 * (fs / cutoff))
     fir_coeff = design_fir_filter(cutoff, fs, numtaps)
     fir_coeff = fir_coeff.to(data.device)
 
     if data.ndim == 1:
-        # 1D 数据处理 - 添加维度以支持反射填充
         data_2d = data.view(1, 1, -1)
         padded_data = torch.nn.functional.pad(data_2d, (numtaps-1, 0), mode='reflect')
         filtered = torch.nn.functional.conv1d(
@@ -34,19 +27,14 @@ def apply_filter(data: torch.Tensor, fs: float, cutoff: float) -> torch.Tensor:
         return filtered.view(-1)
 
     elif data.ndim == 3:
-        # 3D 数据处理
         step, iterations, nrx = data.shape
-        # 重塑数据以使用批量处理
         reshaped_data = data.permute(0, 2, 1).reshape(-1, 1, iterations)
-        # 填充数据
         padded_data = torch.nn.functional.pad(reshaped_data, (numtaps-1, 0), mode='reflect')
-        # 应用卷积
         filtered = torch.nn.functional.conv1d(
             padded_data,
             fir_coeff.view(1, 1, -1),
             padding=0
         )
-        # 重塑回原始形状
         return filtered.view(step, nrx, iterations).permute(0, 2, 1)
 
     else:

deepgpr-0.0.5/src/DeepGPR/wavelet.py

@@ -0,0 +1,34 @@
+from typing import Optional
+import torch
+import math
+
+
+def ricker(
+    freq: float,
+    length: int,
+    dt: float,
+    peak_time: float,
+    dtype: Optional[torch.dtype] = None,
+) -> torch.Tensor:
+    """Return a Ricker wavelet with the specified central frequency.
+
+    Args:
+        freq: The central frequency.
+        length: The number of time samples.
+        dt: The time sample spacing.
+        peak_time: The time (in secs) of the peak amplitude.
+        dtype: The PyTorch datatype to use. Optional, defaults to PyTorch's
+            default (float32).
+    Returns:
+        A PyTorch tensor representing the Ricker wavelet.
+    """
+    if dt == 0:
+        raise ValueError("dt cannot be zero.")
+
+    t: torch.Tensor = torch.arange(float(length), dtype=dtype) * dt - peak_time
+    y: torch.Tensor = (1 - 2 * math.pi**2 * freq**2 * t**2) * torch.exp(
+        -(math.pi**2) * freq**2 * t**2,
+    )
+    if dtype is not None:
+        return y.to(dtype)
+    return y

{deepgpr-0.0.2 → deepgpr-0.0.5/src/DeepGPR.egg-info}/PKG-INFO

@@ -1,10 +1,11 @@
 Metadata-Version: 2.4
 Name: DeepGPR
-Version: 0.0.2
+Version: 0.0.5
 Summary: PyTorch and CUDA for GPR FWI
 Author-email: Lei Liu <liulei990222@gmail.com>
 Classifier: Programming Language :: Python :: 3
-Classifier: Operating System :: OS Independent
+Classifier: Operating System :: Microsoft :: Windows
+Classifier: Operating System :: POSIX :: Linux
 Requires-Python: >=3.7
 Description-Content-Type: text/markdown
 
@@ -59,7 +60,6 @@ peak_time = 1 / freq
 source_amplitudes = torch.zeros((1,nt,1),device=device)
 source_amplitudes[0,:,0]=DeepGPR.ricker(freq, nt, dt, peak_time).to(device)
 
-DeepGPR.plot_survey_geometry(er,source_location, receiver_location,dx)
 
 #forward modeling
 r = DeepGPR.compute(
@@ -79,8 +79,7 @@ ax[1].imshow(er.grad.detach())
 ax[1].set_title("Gradient")
 plt.show()
 ```
-![result](./Fig/example1.png)
 
-![result](./Fig/example2.png)
+![result](./Fig/example.png)
 
 There are more examples in the ./examples.

deepgpr-0.0.5/src/DeepGPR.egg-info/SOURCES.txt

@@ -0,0 +1,13 @@
+README.md
+pyproject.toml
+src/DeepGPR/__init__.py
+src/DeepGPR/common.py
+src/DeepGPR/compute2.py
+src/DeepGPR/multiscale.py
+src/DeepGPR/wavelet.py
+src/DeepGPR.egg-info/PKG-INFO
+src/DeepGPR.egg-info/SOURCES.txt
+src/DeepGPR.egg-info/dependency_links.txt
+src/DeepGPR.egg-info/top_level.txt
+src/DeepGPR/lib/deepgpr.cu
+src/DeepGPR/lib/deepgpr.so

deepgpr-0.0.2/DeepGPR/visual.py

@@ -1,105 +0,0 @@
-import torch
-import numpy as np
-import matplotlib.pyplot as plt
-
-
-def plot_survey_geometry(model, sources, receivers, dx=0.02, title="Survey Geometry"):
-
-    # 1. 数据转换与展平
-    if isinstance(model, torch.Tensor):
-        model = model.detach().cpu().numpy()
-    if isinstance(sources, torch.Tensor):
-        sources = sources.detach().cpu().numpy()
-    if isinstance(receivers, torch.Tensor):
-        receivers = receivers.detach().cpu().numpy()
-
-    # 处理多炮维度 [Shots, N, 3] -> [Total, 3]
-    if sources.ndim == 3: sources = sources.reshape(-1, 3)
-    if receivers.ndim == 3: receivers = receivers.reshape(-1, 3)
-
-    nx, ny, nz = model.shape
-    
-    # 2. 绘图
-    if nz == 1:
-        # ==========================================
-        # 2D 绘图模式 (垂直朝上，0在顶部)
-        # ==========================================
-        fig, ax = plt.subplots(figsize=(6, 4)) 
-        
-        slice_data = model[:, :, 0] # Shape: (nx, ny)
-        
-        # [修改 1] extent 乘以 dx，转换为真实物理距离
-        # extent=[左, 右, 下, 上] 
-        # 垂直轴从 nx*dx (底部) 到 0 (顶部)
-        extent_real = [0, ny * dx, nx * dx, 0]
-        
-        im = ax.imshow(slice_data, cmap='jet', origin='upper', 
-                       extent=extent_real, aspect='auto')
-        
-        plt.colorbar(im, ax=ax, label='Relative permittivity')
-
-        # [修改 2] 散点坐标乘以 dx
-        # Scatter x (水平) = sources[:, 1] * dx
-        # Scatter y (垂直) = sources[:, 0] * dx
-        ax.scatter(sources[:, 1] * dx, sources[:, 0] * dx, c='red', marker='*', s=150, 
-                   label='Sources', edgecolors='k', zorder=10)
-        
-        ax.scatter(receivers[:, 1] * dx, receivers[:, 0] * dx, c='white', marker='v', s=80, 
-                   label='Receivers', edgecolors='k', zorder=9, alpha=0.7)
-
-        # [修改 3] 标签改为物理单位
-        ax.set_xlabel('Distance (m)') 
-        ax.set_ylabel('Depth (m)') 
-        
-        # 保持 Y 轴方向控制 (0 在上)
-        if ax.get_ylim()[0] < ax.get_ylim()[1]: 
-             ax.invert_yaxis()
-
-        # ax.set_title(f"{title} ")
-        ax.legend(loc='lower right')
-        ax.grid(True, linestyle='--', alpha=0.3)
-
-    else:
-        # ==========================================
-        # 3D 绘图模式 (同步应用 dx 以保持一致)
-        # ==========================================
-        fig = plt.figure(figsize=(12, 10))
-        ax = fig.add_subplot(111, projection='3d')
-
-        # 散点坐标乘以 dx
-        ax.scatter(sources[:, 0] * dx, sources[:, 1] * dx, sources[:, 2] * dx, 
-                   c='red', marker='*', s=100, label='Sources')
-        ax.scatter(receivers[:, 0] * dx, receivers[:, 1] * dx, receivers[:, 2] * dx, 
-                   c='blue', marker='v', s=40, label='Receivers', alpha=0.6)
-
-        # 网格坐标乘以 dx
-        x_grid = np.arange(nx) * dx
-        y_grid = np.arange(ny) * dx
-        z_grid = np.arange(nz) * dx
-        
-        cx, cy, cz = nx // 2, ny // 2, nz // 2
-        
-        # 生成真实坐标的 Meshgrid
-        X, Y = np.meshgrid(x_grid, y_grid, indexing='ij') 
-        # 注意 offset 也要乘以 dx
-        ax.contourf(X, Y, model[:, :, cz], zdir='z', offset=cz * dx, cmap='viridis', alpha=0.5)
-        
-        X, Z = np.meshgrid(x_grid, z_grid, indexing='ij')
-        ax.contourf(X, model[:, cy, :], Z, zdir='y', offset=cy * dx, cmap='viridis', alpha=0.5)
-        
-        Y, Z = np.meshgrid(y_grid, z_grid, indexing='ij')
-        ax.contourf(model[cx, :, :], Y, Z, zdir='x', offset=cx * dx, cmap='viridis', alpha=0.5)
-
-        # 设置真实坐标范围
-        ax.set_xlim(0, nx * dx)
-        ax.set_ylim(0, ny * dx)
-        ax.set_zlim(0, nz * dx)
-        
-        ax.set_xlabel('X (m)')
-        ax.set_ylabel('Y (m)')
-        ax.set_zlabel('Z (m)')
-        ax.set_title(title)
-        ax.legend()
-
-    plt.tight_layout()
-    plt.show()

deepgpr-0.0.2/DeepGPR.egg-info/SOURCES.txt

@@ -1,13 +0,0 @@
-README.md
-pyproject.toml
-DeepGPR/__init__.py
-DeepGPR/common.py
-DeepGPR/compute2.py
-DeepGPR/multiscale.py
-DeepGPR/visual.py
-DeepGPR.egg-info/PKG-INFO
-DeepGPR.egg-info/SOURCES.txt
-DeepGPR.egg-info/dependency_links.txt
-DeepGPR.egg-info/top_level.txt
-DeepGPR/lib/deepgpr.cu
-DeepGPR/lib/deepgpr.so