refactor(demo): 重构示例脚本目录结构

scripts/download.py

@@ -1,14 +0,0 @@
-import os
-from huggingface_hub import snapshot_download
-
-
-PROJECT_ROOT = os.path.dirname(
-    os.path.dirname(os.path.abspath(__file__)))
-
-
-if __name__ == "__main__":
-    snapshot_download(
-        repo_id="ViperEk/KHAOSZ",
-        local_dir=os.path.join(PROJECT_ROOT, "params"),  
-        force_download=True
-    )

scripts/generate_ar.py

@@ -1,27 +0,0 @@
-import os
-import torch
-from khaosz import Khaosz
-
-
-PROJECT_ROOT = os.path.dirname(
-    os.path.dirname(os.path.abspath(__file__)))
-
-def generate_text():
-    model_dir = os.path.join(PROJECT_ROOT, "params")
-    model = Khaosz(model_dir).to(device='cuda', dtype=torch.bfloat16)
-
-    query = input(">> ")
-    
-    response = model.text_generate(
-        query=query, 
-        temperature=0.8,
-        top_p=0.95,
-        top_k=50
-    )
-    
-    print(response)
-
-    
-
-if __name__ == "__main__":
-    generate_text()

scripts/generate_batch.py

@@ -1,25 +0,0 @@
-import os
-import torch
-from khaosz import Khaosz
-
-
-PROJECT_ROOT = os.path.dirname(
-    os.path.dirname(os.path.abspath(__file__)))
-
-def batch_generate():
-    model_dir = os.path.join(PROJECT_ROOT, "params")
-    model = Khaosz(model_dir).to(device='cuda', dtype=torch.bfloat16)
-    inputs = ["你好", "请问什么是人工智能", "今天天气如何", "我感到焦虑， 请问我应该怎么办", "请问什么是显卡"]
-    
-    responses = model.batch_generate(
-        queries=inputs,
-        temperature=0.8,
-        top_p=0.95,
-        top_k=50
-    )
-    
-    for q, r in zip(inputs, responses):
-        print((q, r))
-
-if __name__ == "__main__":
-    batch_generate()

scripts/generate_retrieve.py

@@ -1,42 +0,0 @@
-import os
-import torch
-from khaosz import Khaosz, SemanticTextSplitter, Retriever
-
-
-PROJECT_ROOT = os.path.dirname(
-    os.path.dirname(os.path.abspath(__file__)))
-
-if __name__ == "__main__":
-    model_dir = os.path.join(PROJECT_ROOT, "params")
-    context_path = os.path.join(PROJECT_ROOT, "README.md")
-    
-    model = Khaosz(model_dir).to(device='cuda', dtype=torch.bfloat16)
-    spliter = SemanticTextSplitter(model.encode)
-    retriever = Retriever()
-    text = open(context_path, "r", encoding="utf-8").read()
-    
-    res = spliter.split(text, threshold=0.8, window_size=1)
-    # print(("\n" + "+"*100 + "\n").join(res))
-
-    res_embs = model.encode(res)
-    for sentence, emb in zip(res, res_embs):
-        retriever.add_vector(sentence, emb)
-
-    retrive_top_k = 5
-    query = "作者设计了一个怎样的模型"
-    emb_query = model.encode(query)
-    retrieved = retriever.retrieve(emb_query, retrive_top_k)
-    
-    retrive_response = model.retrieve_generate(
-        retrieved=retrieved,
-        query=query,
-        temperature=0.8,
-        top_p=0.95,
-        top_k=50
-    )
-
-    print("retrive content:")
-    print("\n".join([f"{idx + 1}. " + text for idx, (text, _) in enumerate(retrieved)]))
-
-    print("\n\nretrive generate:")
-    print(retrive_response)

scripts/stream_chat.py

@@ -1,32 +0,0 @@
-import os
-import torch
-from khaosz import Khaosz
-
-
-PROJECT_ROOT = os.path.dirname(
-    os.path.dirname(os.path.abspath(__file__)))
-
-def chat():
-    model_dir = os.path.join(PROJECT_ROOT, "params")
-    model = Khaosz(model_dir).to(device='cuda', dtype=torch.bfloat16)
-
-    history = []
-    while True:
-        query = input(">> ")
-        if query == "!exit":
-            break
-        
-        response_size = 0
-        for response, history in model.stream_generate(
-            query=query, 
-            history=history,
-            temperature=0.8,
-            top_p=0.95,
-            top_k=50
-        ):
-            print(response[response_size:], end="", flush=True)
-            response_size = len(response)
-    
-
-if __name__ == "__main__":
-    chat()