Merge branch 'main' of github.com:OpenBMB/ChatDev

ecl/ece.py

@@ -0,0 +1,146 @@
+
+import os
+import json
+import re
+import numpy as np
+import argparse
+point = 0.95
+eliminate_threshold = 0.95
+
+
+def retrieve_eliminate(Path_directory,UsedMemory_directory,Evolved_directory):
+    experiences_use = []
+    content = []
+    content1 = []
+    experiences_total = []
+    usetime_total = []
+    exp_dict = {}
+    eliminated_exp = []
+
+    directories = [os.path.join(Path_directory, d) for d in os.listdir(Path_directory) if os.path.isdir(os.path.join(Path_directory, d))]
+    for subdir in directories:
+        directory = subdir
+        logdir = [filename for filename in os.listdir(directory) if filename.endswith(".log")]
+        logdir = os.path.join(directory, logdir[0])
+        content1 = open(logdir, "r", encoding='UTF-8').read()
+
+        pattern1 = re.compile(r'the source code MIDs is (.*?),', re.S)
+        experiences_sourceMIDs = re.findall(pattern1, content1)
+        pattern2 = re.compile(r'the target code MIDs is (.*?)\n',re.S)
+        experiences_targetMIDs = re.findall(pattern2, content1)
+        pattern3 = re.compile(r'And the (.*?) similarity is',re.S)
+        experiences_type = re.findall(pattern3,content1)
+        for i in range(0,len(experiences_sourceMIDs)):
+            sourceMID = experiences_sourceMIDs[i]
+            targetMID = experiences_targetMIDs[i]
+            type = experiences_type[i]
+            experiences_use.append((sourceMID,targetMID,type))
+
+    with open(UsedMemory_directory) as file:
+        content1 = json.load(file)
+        new_content = []
+        for memorypiece in content1:
+            experiences = memorypiece.get("experiences")
+            if experiences != None:
+                experiences_total.extend(experiences)
+                for experience in experiences:
+                    experience["use_time"] = 0
+        for experience in experiences_use:
+            for experience_t in experiences_total:
+                if experience[0] == experience_t["sourceMID"] and experience[1] == experience_t["targetMID"]:
+                    experience_t["use_time"] += 1
+        for i,experience_t in enumerate(experiences_total):
+            usetime_total.append(experience_t["use_time"])
+            exp_dict[i] = experience_t["use_time"]
+        file.close()
+
+    usetime_sort = sorted(usetime_total)[::-1]
+    total = np.sum(usetime_sort)
+    for i in range(len(usetime_sort)):
+        if np.sum(usetime_sort[:i])/total >= point:
+            # print("α：",i)
+            alpha= i
+            break
+    index=0
+    for k in sorted(exp_dict,key=exp_dict.__getitem__,reverse=True):
+        if index <= alpha:
+            eliminated_exp.append(experiences_total[k])
+            index += 1
+        else:
+            break
+
+    for memorypiece in content1:
+        experiences = memorypiece.get("experiences")
+        retrieve_eliminated_experienceList = []
+        if experiences != None:
+            for experience in experiences:
+                if experience in eliminated_exp:
+                    retrieve_eliminated_experienceList.append(experience)
+
+        memorypiece["experiences"] = retrieve_eliminated_experienceList
+        new_content.append(memorypiece)
+
+    with open(Evolved_directory, 'w') as file:
+        json.dump(new_content, file)
+
+
+# Quality score gain Elimination
+def gain_eliminate(NewMemory_directory,Evolved_directory):
+    content2 = []
+    with open(NewMemory_directory) as file:
+        content2 = json.load(file)
+        new_content2 = []
+        for memorypiece in content2:
+            experiences = memorypiece.get("experiences")
+            gain_eliminated_experienceList = []
+
+            if experiences != None:
+                # print("origin:", len(experiences))
+                for experience in experiences:
+                    valueGain = experience.get("valueGain")
+                    # print(valueGain)
+                    if valueGain >= eliminate_threshold:
+                        gain_eliminated_experienceList.append(experience)
+                # print(len(experiences))
+                memorypiece["experiences"] = gain_eliminated_experienceList
+                new_content2.append(memorypiece)
+            else:
+                new_content2.append(memorypiece)
+        file.close()
+
+    with open(Evolved_directory, 'r') as file:
+        new_content = json.load(file)
+
+    new_content = new_content + new_content2
+
+    with open(Evolved_directory, 'w') as file:
+        json.dump(new_content, file)
+
+
+
+def recount_experience(Evolved_directory):
+    with open(Evolved_directory, 'r') as file:
+        content = json.load(file)
+
+    with open(Evolved_directory, 'w') as file:
+        i = 0
+        for memorypiece in content:
+            memorypiece["total"] = i
+            i += 1
+        json.dump(content, file)
+
+def main():
+    parser = argparse.ArgumentParser(description="Process memory with some directories.")
+    parser.add_argument("Path_directory", type = str, help="The directory of software")
+    parser.add_argument("UsedMemory_directory", type=str, help="The directory of MemoryCards")
+    parser.add_argument("NewMemory_directory", type=str, help="The directory of NewMemoryCards")
+    parser.add_argument("Evolved_directory", type= str, help="The directory for output")
+
+
+    args = parser.parse_args()
+    retrieve_eliminate(args.Path_directory,args.UsedMemory_directory,args.Evolved_directory)
+    gain_eliminate(args.NewMemory_directory,args.Evolved_directory)
+    recount_experience(args.Evolved_directory)
+
+if __name__ == "__main__":
+    main()

wiki.md

@@ -149,7 +149,7 @@ then start building a software by ``python3 run.py`` and go to [Visualizer Websi
 After this process, the experiences have been extracted from the production of software and added to the agents' experience pool in `ecl/memory/MemoryCards.json`.
 \
 **For example:**
-  It you want to memorize only one software, you can use:
+  If you want to memorize only one software, you can use:
   ```bash
     python3 ecl/ecl.py "<Software Path to file>"
   ```
@@ -187,6 +187,25 @@ After this process, the experiences have been extracted from the production of s
 
 Detailed descriptions and experiment results about this **Experiential Co-Learning** Module lies in our preprint paper at https://arxiv.org/abs/2312.17025.
 
+## Experiential Co-Evolving Guide
+- **Using Co-Evolving**: Use the following command to initiate the evolving of experiences, which uses the `ecl/ece.py` to eliminate `ecl/memory/UsedMemory.json` and `ecl/memory/NewMemory.json`. Then it combines the two parts of experiences to form a new experience pool in `ecl/memory/Evolved_directory.json`.
+
+  ```bash
+    python3 ecl/ece.py "<Path_directory>" "<UsedMemory_directory>" "<NewMemory_directory>" "<Evolved_directory>"
+  ```
+  `<Path_directory>`: The path to the directory of software , generated with the memory `UsedMemory_directory`. \
+  `<UsedMemory_directory>`: The path to the directory of UsedMemory, which was used to generate the software in `Path_directory`. \
+  `<NewMemory_directory>`: The path to the directory NewMemory, which acquires from the software in `Path_directory` using `ecl/ecl.py`. \
+  `<Evolved_directory>`: The path to a directory where you want to store the evolved memory.
+  \
+  **For example:**
+  ```bash
+    python3 ecl/ece.py "WareHouse" "ecl/memory/UsedMemory.json" "ecl/memory/NewMemory.json" "ecl/memory/MemoryCards_Evolved.json"
+  ```
+> **Notice:** The software directory and memory directory must correspond. The software in the "<Path_directory>" is generated using "<UsedMemory_directory>", and the "<NewMemory_directory>" is acquired from the software in the  "<Path_directory>". That's because when we calculate the frequency distribution of the experience, we need to ensure that the software corresponds to the experiences, which to eliminate certain experiences to obtain a subset with relatively high retrieval probability.
+
+Detailed descriptions and experiment results about this Experiential Co-Evolving Module lies in our preprint paper at https://arxiv.org/abs/2405.04219.
+
 ## Customization
 
 - You can customize your company in three kinds of granularity: