joonmyung 1.5.9__tar.gz → 1.5.11__tar.gz

{joonmyung-1.5.9 → joonmyung-1.5.11}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: joonmyung
-Version: 1.5.9
+Version: 1.5.11
 Summary: JoonMyung's Library
 Home-page: https://github.com/pizard/JoonMyung.git
 Author: JoonMyung Choi

{joonmyung-1.5.9 → joonmyung-1.5.11}/joonmyung/analysis/analysis.py

@@ -14,17 +14,12 @@ import cv2
 
 def anaModel(transformer_class):
     class VisionTransformer(transformer_class):
-        def has_parameter(self, parameter_name):
-            return parameter_name in self.__init__.__code__.co_varnames
-
         def forward_features(self, x):
             x = self.patch_embed(x)
-            if self.has_parameter("cls_token"):
+            if hasattr(self, "cls_token"):
                 cls_token = self.cls_token.expand(x.shape[0], -1, -1)  # stole cls_tokens impl from Phil Wang, thanks
                 x = torch.cat((cls_token, x), dim=1)
 
-
-
             if self.analysis[0] == 1:   # PATCH
                 x = x # (8, 197, 192)
             elif self.analysis[0] == 2: # POS
@@ -39,27 +34,27 @@ def anaModel(transformer_class):
 
             x = self.blocks(x)
             x = self.norm(x)
-            if self.has_parameter("cls_token") and self.has_parameter("dist_token"):
+            if hasattr(self, "cls_token") and hasattr(self, "cls_token"):
                 return x[:, 0], x[:, 1]
-            elif self.has_parameter("cls_token"):
+            elif hasattr(self, "cls_token"):
                 return self.pre_logits(x[:, 0])
             else:
                 return self.pre_logits(x.mean(dim=1))
 
-
     return VisionTransformer
 
 class Analysis:
     def __init__(self, model, analysis = [0], activate = [True, False, False, False], detach=True, key_name=None, num_classes = 1000
-                 , cls_start=0, cls_end=1, patch_start=1, patch_end=None
+                 , cls_start=0, cls_end=1, patch_start=1, patch_end=None, wrapping=False
                  , amp_autocast=suppress, device="cuda"):
         # Section A. Model
         self.num_classes = num_classes
         self.key_name = key_name
+        if wrapping:
+            model_ = anaModel(model.__class__)
+            model.__class__ = model_
+            model.analysis = analysis
 
-        model_ = anaModel(model.__class__)
-        model.__class__ = model_
-        model.analysis = analysis
         self.model = model
         self.detach = detach
 
@@ -72,8 +67,7 @@ class Analysis:
                  {"name_i": 'qkv', "name_o": 'decoder', "fn_f": self.qkv_forward, "fn_b": self.qkv_backward},
                  {"name_i": 'head', "name_o": 'decoder', "fn_f": self.head_forward, "fn_b": self.head_backward},
                  {"name_i": 'patch_embed.norm', "name_o": 'decoder', "fn_f": self.input_forward, "fn_b": self.input_backward}]
-        hooks = [h for h, a in zip(hooks, activate) if a]
-
+        self.activate = activate
 
         self.amp_autocast = amp_autocast
         self.device       = device
@@ -87,15 +81,15 @@ class Analysis:
 
     def attn_forward(self, module, input, output):
         # input/output : 1 * (8, 3, 197, 197) / (8, 3, 197, 197)
-        self.info["attn"]["f"] = output.detach() if self.detach else output
+        if self.activate[0]: self.info["attn"]["f"] = output.detach() if self.detach else output
 
     def attn_backward(self, module, grad_input, grad_output):
         # input/output : 1 * (8, 3, 197, 192) / (8, 3, 197, 576)
-        self.info["attn"]["b"] = grad_input[0].detach() if self.detach else grad_input[0]
+        if self.activate[0]: self.info["attn"]["b"] = grad_input[0].detach() if self.detach else grad_input[0]
 
     def qkv_forward(self, module, input, output):
         # input/output : 1 * (8, 197, 192) / (8, 197, 576)
-        self.info["qkv"]["f"].append(output.detach())
+        if self.activate[1]: self.info["qkv"]["f"].append(output.detach())
 
     def qkv_backward(self, module, grad_input, grad_output):
         # self.info["qkv"]["b"].append(grad_input[0].detach())
@@ -103,20 +97,22 @@ class Analysis:
 
     def head_forward(self, module, input, output):
         # input : 1 * (8(B), 192(D)), output : (8(B), 1000(C))
-        B = output.shape[0]
-        pred = targetPred(output, self.targets, topk=5)
-        self.info["head"]["TF"] += (pred[:, 0] == pred[:, 1])
+        if self.activate[2]:
+            B = output.shape[0]
+            pred = targetPred(output, self.targets, topk=5)
+            self.info["head"]["TF"] += (pred[:, 0] == pred[:, 1])
 
-        acc1, acc5 = accuracy(output, self.targets, topk=(1,5))
-        self.info["head"]["acc1"].update(acc1.item(), n=B)
-        self.info["head"]["acc5"].update(acc5.item(), n=B)
+            acc1, acc5 = accuracy(output, self.targets, topk=(1,5))
+            self.info["head"]["acc1"].update(acc1.item(), n=B)
+            self.info["head"]["acc5"].update(acc5.item(), n=B)
 
     def head_backward(self, module, grad_input, grad_output):
         pass
 
     def input_forward(self, module, input, output):
-        norm = F.normalize(output, dim=-1)
-        self.info["input"]["sim"] += (norm @ norm.transpose(-1, -2)).mean(dim=(-1, -2))
+        if self.activate[3]:
+            norm = F.normalize(output, dim=-1)
+            self.info["input"]["sim"] += (norm @ norm.transpose(-1, -2)).mean(dim=(-1, -2))
 
     def input_backward(self, module, grad_input, grad_output):
         pass

{joonmyung-1.5.9 → joonmyung-1.5.11}/joonmyung/script.py

@@ -97,7 +97,7 @@ class GPU_Worker():
         return
 
     def message(self, text):
-        url = "https://hooks.slack.com/services/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"
+        url = "https://hooks.slack.com/services/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"
         payload = {"text": text}
         headers = {'Content-type': 'application/json'}
 
@@ -123,7 +123,7 @@ def Process_Worker(processes, gpuWorker, id = "", p = True):
     training_time = datetime.timedelta(seconds=time.mktime(end) - time.mktime(start))
     print(f"Time 1/all :  {training_time}/{training_time / len(processes)} ------")
     gpuWorker.message(f"Experiments Finished" 
-                      f"{id} : "
+                      f"{id} : {server}"
                       f"Time 1/all : {training_time}/{training_time / len(processes)}"
                       )
 

{joonmyung-1.5.9 → joonmyung-1.5.11}/joonmyung.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: joonmyung
-Version: 1.5.9
+Version: 1.5.11
 Summary: JoonMyung's Library
 Home-page: https://github.com/pizard/JoonMyung.git
 Author: JoonMyung Choi

{joonmyung-1.5.9 → joonmyung-1.5.11}/setup.py

@@ -3,7 +3,7 @@ from setuptools import find_packages
 
 setuptools.setup(
     name="joonmyung",
-    version="1.5.9",
+    version="1.5.11",
     author="JoonMyung Choi",
     author_email="pizard@korea.ac.kr",
     description="JoonMyung's Library",