RAG\u57fa\u672c\u4ecb\u7d39<\/h2>\n\n\n\n
RAG \u6280\u8853\u662f\u4e00\u7a2e\u7d50\u5408\u4e86\u641c\u5c0b\u6aa2\u7d22\u548c\u751f\u6210\u80fd\u529b\u7684\u81ea\u7136\u8a9e\u8a00\u8655\u7406\u67b6\u69cb\uff0c\u5b83\u53ef\u4ee5\u5f9e\u5916\u90e8\u77e5\u8b58\u5eab\u641c\u5c0b\u76f8\u95dc\u4fe1\u606f\uff0c\u7136\u5f8c\u4f7f\u7528\u9019\u4e9b\u4fe1\u606f\u4f86\u751f\u6210\u56de\u61c9\u6216\u5b8c\u6210\u7279\u5b9a\u7684 NLP \u4efb\u52d9\u3002<\/p>\n\n\n\n
RAG \u6a21\u578b\u4e3b\u8981\u7531\u5169\u500b\u90e8\u5206\u69cb\u6210\uff1a\u6aa2\u7d22\u5668\u548c\u751f\u6210\u5668\u3002<\/p>\n\n\n\n
- \n
- \u6aa2\u7d22\u5668\u8ca0\u8cac\u5f9e\u5916\u90e8\u77e5\u8b58\u5eab\uff08\u4f8b\u5982\uff0c\u6587\u672c\u6578\u64da\u5eab\u6216\u9810\u5148\u8a13\u7df4\u7684\u77e5\u8b58\u5d4c\u5165\uff09\u4e2d\u6aa2\u7d22\u76f8\u95dc\u7684\u77e5\u8b58\u8a0a\u606f<\/li>\n\n\n\n
- \u751f\u6210\u5668\u6703\u5229\u7528\u6aa2\u7d22\u5230\u7684\u77e5\u8b58\u4f86\u751f\u6210\u56de\u61c9\u3002<\/li>\n<\/ol>\n\n\n\n
\u900f\u904e\u9019\u500b\u67b6\u69cb\uff0c\u6a21\u578b\u53ef\u4ee5\u5f9e\u5927\u91cf\u7684\u6587\u672c\u6578\u64da\u4e2d\u627e\u5230\u8207\u554f\u984c\u76f8\u95dc\u7684\u7b54\u6848\uff0c\u4e26\u751f\u6210\u7d50\u69cb\u5316\u7684\u56de\u61c9\u3002RAG \u6280\u8853\u7684\u512a\u9ede\u662f\u5229\u7528\u5927\u91cf\u7684\u771f\u5be6\u6578\u64da\uff0c\u5f97\u51fa\u66f4\u6e96\u78ba\u53ca\u66f4\u5177\u6709\u8a9e\u5883\u7684\u56de\u7b54\u3002<\/p>\n\n\n\n
\u70ba\u4ec0\u9ebc\u9700\u8981 RAG<\/h2>\n\n\n\n
\u5b83\u4e3b\u8981\u89e3\u6c7a\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u8655\u7406\u5373\u6642\u8a0a\u606f\uff0c\u4ee5\u53ca\u66f4\u65b0\u8cc7\u8a0a\u7684\u9650\u5236:<\/p>\n\n\n\n
- \n
- \u8655\u7406\u5373\u6642\u6216\u6700\u65b0\u4fe1\u606f\uff1a<\/strong> \u50b3\u7d71\u7684\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u5728\u7279\u5b9a\u6642\u9593\u9ede\u7684\u6578\u64da\u5feb\u7167\u4e0a\u9032\u884c\u8a13\u7df4\uff0c\u56e0\u6b64\u5b83\u5011\u50c5\u80fd\u56de\u7b54\u8a13\u7df4\u6578\u64da\u4e2d\u5305\u542b\u7684\u4fe1\u606f\u3002\u4f8b\u5982\uff0c\u554f\u4e00\u500b LLM \u4eca\u5929\u7684\u5929\u6c23\u5982\u4f55\uff0c\u5b83\u53ef\u80fd\u7121\u6cd5\u56de\u7b54\uff0c\u56e0\u70ba\u4eca\u5929\u7684\u5929\u6c23\u4fe1\u606f\u4e0d\u5305\u542b\u5728\u5b83\u7684\u8a13\u7df4\u6578\u64da\u4e2d\u3002RAG \u900f\u904e\u63a2\u7d22\u5373\u6642\u5929\u6c23\u7684 API \u6216\u76f8\u95dc\u4fe1\u606f\u4f86\u7372\u5f97\u8cc7\u8a0a\uff0c\u4e26\u5728\u95b1\u8b80\u9019\u4e9b API \u7684\u7d50\u679c\u5f8c\uff0c\u751f\u6210\u9700\u8981\u7684\u7b54\u6848\uff0c\u4ee5\u5f4c\u88dc\u8655\u7406\u5373\u6642\u554f\u984c\u7684\u9650\u5236\u3002<\/li>\n\n\n\n
- \u63d0\u4f9b\u66f4\u6e96\u78ba\u548c\u76f8\u95dc\u7684\u56de\u7b54\uff1a<\/strong> \u901a\u904e\u6aa2\u7d22\u548c\u6574\u5408\u8207\u7528\u6236\u67e5\u8a62\u76f8\u95dc\u7684\u5916\u90e8\u4fe1\u606f\uff0cRAG \u80fd\u5920\u4f7f\u6a21\u578b\u63d0\u4f9b\u66f4\u6e96\u78ba\u3001\u66f4\u5177\u76f8\u95dc\u6027\u7684\u7b54\u6848\u3002\u9019\u5c0d\u65bc\u9700\u8981\u5c08\u696d\u77e5\u8b58\u6216\u7279\u5b9a\u9818\u57df\u4fe1\u606f\u7684\u67e5\u8a62\u7279\u5225\u6709\u7528\uff0c\u9019\u4e9b\u5c08\u696d\u77e5\u8b58\u53ef\u80fd\u4f86\u81ea\u67d0\u4e9b\u9577\u7bc7\u7684\u6587\u4ef6\u4e2d\uff0c\u90a3\u5c31\u5f88\u9069\u5408\u900f\u904e RAG \u53bb\u6aa2\u7d22\u8207\u751f\u6210\uff0c\u4e14\u751f\u6210\u7684\u5167\u5bb9\u4e5f\u6703\u6839\u64da\u6b64\u9577\u6587\uff0c\u5c0e\u51fa\u7cbe\u78ba\u7684\u56de\u7b54\u3002<\/li>\n\n\n\n
- \u64f4\u5c55\u6a21\u578b\u7684\u61c9\u7528\u7bc4\u570d\uff1a<\/strong> RAG \u901a\u904e\u6574\u5408\u4e0d\u540c\u7684\u6578\u64da\u6e90\uff08\u5982\u5be6\u6642API\u6578\u64da\u3001\u5c08\u696d\u6587\u4ef6\u7b49\uff09\uff0c\u4f7f\u6a21\u578b\u80fd\u5920\u89e3\u7b54\u539f\u672c\u8d85\u51fa\u5176\u8a13\u7df4\u7bc4\u570d\u7684\u554f\u984c\uff0c\u5f9e\u800c\u64f4\u5c55\u4e86\u6a21\u578b\u7684\u61c9\u7528\u7bc4\u570d\u3002<\/li>\n\n\n\n
- \u512a\u5316\u6210\u672c\u548c\u6548\u7387\uff1a<\/strong> \u5c0d\u65bc\u9700\u8981\u8655\u7406\u5927\u91cf\u6587\u6a94\u6216\u6578\u64da\u7684\u5834\u666f\uff0cRAG \u53ef\u4ee5\u901a\u904e\u53ea\u6aa2\u7d22\u548c\u8655\u7406\u8207\u7279\u5b9a\u67e5\u8a62\u6700\u76f8\u95dc\u7684\u4fe1\u606f\uff0c\u4f86\u63d0\u9ad8\u8655\u7406\u6548\u7387\u548c\u964d\u4f4e\u904b\u7b97\u6210\u672c\u3002<\/li>\n<\/ol>\n\n\n\n
\u8a9e\u7fa9\u6aa2\u7d22\uff08Semantic Search\uff09\u57fa\u672c\u4ecb\u7d39<\/h2>\n\n\n\n
\u8207RAG\u7dca\u5bc6\u76f8\u9023\u7684\u53e6\u4e00\u9805\u95dc\u9375\u6280\u8853\u662f\u8a9e\u7fa9\u6aa2\u7d22\u3002\u8a9e\u7fa9\u6aa2\u7d22\u6280\u8853\u4e3b\u8981\u7528\u65bc\u5f9e\u5927\u898f\u6a21\u6578\u64da\u96c6\u4e2d\u5feb\u901f\u8b58\u5225\u51fa\u8207\u7279\u5b9a\u67e5\u8a62\u8a9e\u7fa9\u4e0a\u6700\u76f8\u95dc\u7684\u4fe1\u606f\u7247\u6bb5<\/strong>\u3002\u9019\u7a2e\u6280\u8853\u901a\u904e\u5c07\u6587\u672c\u8f49\u63db\u70ba\u6578\u503c\u5411\u91cf\uff08\u5d4c\u5165\uff09\uff0c\u4e26\u8a08\u7b97\u9019\u4e9b\u5411\u91cf\u4e4b\u9593\u7684\u76f8\u4f3c\u5ea6\uff0c\u4f86\u78ba\u5b9a\u54ea\u4e9b\u6587\u672c\u7247\u6bb5\u8207\u7528\u6236\u7684\u67e5\u8a62\u6700\u70ba\u5339\u914d\u3002<\/p>\n\n\n\n
\u5982\u679c\u60f3\u8b93\u6a21\u578b\u56de\u7b54\u95dc\u65bc\u4e00\u672c\u6f5b\u6c34\u8247\u624b\u518a\uff08\u9577\u9054492\u9801\uff09\u7684\u554f\u984c\uff0c\u76f4\u63a5\u5c07\u6574\u672c\u624b\u518a\u4f5c\u70ba\u4e0a\u4e0b\u6587\u63d0\u4f9b\u7d66\u6a21\u578b\u662f\u4e0d\u5207\u5be6\u969b\u7684\u3002\u9019\u662f\u56e0\u70ba\u9019\u6a23\u7684\u5927\u91cf\u6587\u672c\u6703\u8d85\u904e\u6a21\u578b\u7684\u8655\u7406\u80fd\u529b\u548c\u6700\u5927\u6a19\u8a18\u9650\u5236\u3002\u6b64\u5916\uff0c\u7531\u65bc\u57fa\u65bc\u4f7f\u7528\u7684 token \u6578\u8a08\u8cbb\uff0c\u904e\u9577\u7684\u63d0\u793a\u6703\u5c0e\u81f4\u6210\u672c\u589e\u52a0\u3002<\/p>\n\n\n\n
\u9019\u7a2e\u60c5\u6cc1\u4e0b\uff0c\u8a9e\u7fa9\u6aa2\u7d22\uff08Semantic Search\uff09\u6280\u8853<\/strong>\u5c31\u986f\u5f97\u975e\u5e38\u91cd\u8981\u3002\u5176\u6838\u5fc3\u601d\u60f3\u662f\u5c07\u5927\u578b\u6587\u6a94\u5206\u89e3\u6210\u8f03\u5c0f\u7684\u6587\u672c\u7247\u6bb5\u3002\u9019\u6a23\u4e0d\u50c5\u4f7f\u5f97\u6587\u672c\u8655\u7406\u548c\u6aa2\u7d22\u8b8a\u5f97\u66f4\u9ad8\u6548\uff0c\u4e5f\u4f7f\u6574\u500b\u904e\u7a0b\u66f4\u7d93\u6fdf\u5be6\u7528\u3002<\/p>\n\n\n\n
\u5177\u9ad4\u4f86\u8aaa\uff0c\u6211\u5011\u5c07\u6587\u672c\u7247\u6bb5\u8f49\u63db\u6210 vector\uff0c\u7136\u5f8c\u6e2c\u91cf\u9019\u4e9b\u5411\u91cf\u4e4b\u9593\u7684\u8ddd\u96e2\uff0c\u5f9e\u800c\u8b58\u5225\u51fa\u8207\u7528\u6236\u63d0\u554f\u5728\u8a9e\u7fa9\u4e0a\u6700\u63a5\u8fd1\u7684\u6587\u672c\u7247\u6bb5\u3002\u9019\u6a23\uff0c\u6211\u5011\u5c31\u53ea\u9700\u5c07\u6700\u76f8\u95dc\u7684\u6587\u672c\u90e8\u5206\u4f5c\u70ba\u4e0a\u4e0b\u6587\u63d0\u4f9b\u7d66\u6a21\u578b\uff0c\u800c\u4e0d\u662f\u6574\u500b\u5927\u578b\u6587\u6a94\u3002\u7e3d\u800c\u8a00\u4e4b\uff0c\u900f\u904e\u5c07\u5927\u578b\u6587\u6a94\u5206\u89e3\u4e26\u4f7f\u7528\u8a9e\u7fa9\u6aa2\u7d22\u6280\u8853\uff0c\u6211\u5011\u53ef\u4ee5\u66f4\u9ad8\u6548\u5730\u5229\u7528\u8a9e\u8a00\u6a21\u578b\u4f86\u56de\u7b54\u8907\u96dc\u548c\u5c08\u696d\u7684\u554f\u984c\uff0c\u540c\u6642\u63a7\u5236\u6210\u672c\u548c\u63d0\u5347\u56de\u7b54\u7684\u6e96\u78ba\u6027\u3002<\/p>\n\n\n\n
\u5be6\u4f5c<\/h2>\n\n\n\n
RAG \u56de\u7b54 PDF \u554f\u984c<\/h2>\n\n\n\n
\u5c0e\u5165 pdf \u6587\u4ef6<\/h3>\n\n\n\n
def pdf_to_text(path: str,
start_page: int = 1,
end_page: Optional[int or None] = None) -> list[str]:
\"\"\"
Converts PDF to plain text.
Params:
path (str): Path to the PDF file.
start_page (int): Page to start getting text from.
end_page (int): Last page to get text from.
\"\"\"
loader = PyPDFLoader(path)
pages = loader.load()
total_pages = len(pages)
if end_page is None:
end_page = len(pages)
text_list = []
for i in range(start_page-1, end_page):
text = pages[i].page_content
text = text.replace('\\n', ' ')
text = re.sub(r'\\s+', ' ', text)
text_list.append(text)
return text_list
text_list = pdf_to_text(\"pdfs\/DQ2.pdf\")
print(text_list[:2])<\/pre>\n\n\n\n![\"raw-image\"\/](\"https:\/\/d2a6d2ofes041u.cloudfront.net\/resize?compression=6&norotation=true&url=https%3A%2F%2Fimages.vocus.cc%2F2353c272-4b94-401d-abb5-5e8c08eab2cb.png&width=740&sign=loAUND3IQh_0-NlVBvUlV1qK8CHz3E98fhWQwVeijUU\")
<\/figure>\n<\/div>\n\n\n<\/figure>\n<\/div>\n\n\n<\/figure>\n<\/div>\n\n\n\u6587\u5b57\u9810\u8655\u7406<\/h3>\n\n\n\n
\u4ee5\u9801\u70ba\u55ae\u4f4d\u5206\u5272\uff0c\u4e26\u8a18\u9304\u6bcf\u500b\u7247\u6bb5\u7684\u9801\u78bc\u3002<\/p>\n\n\n\n
def text_to_chunks(texts: list[str],
start_page: int = 1) -> list[list[str]]:
\"\"\"
Splits the text into equally distributed chunks.
Args:
texts (str): List of texts to be converted into chunks.
word_length (int): Maximum number of words in each chunk.
start_page (int): Starting page number for the chunks.
\"\"\"
text_toks = [t.split(' ') for t in texts]
chunks = []
for idx, words in enumerate(text_toks):
chunk = ' '.join(words).strip()
chunk = f'[Page no. {idx+start_page}]' + ' ' + '\"' + chunk + '\"'
chunks.append(chunk)
return chunks
chunks = text_to_chunks(text_list)<\/pre>\n\n\n\n\u8f49\u63db\u6210 vector \u4f7f\u7528 text embeddings<\/h3>\n\n\n\n
\u61c9\u7528 hugging dace \u7684 universal-sentence-encoder_4 <\/strong>\u6a21\u578b\u5c07\u6587\u5b57\u8f49\u63db\u6210 vector<\/p>\n\n\n\n
def get_text_embedding(texts: list[list[str]],
batch: int = 1000) -> list[Any]:
\"\"\"
Get the embeddings from the text.
Args:
texts (list(str)): List of chucks of text.
batch (int): Batch size.
\"\"\"
embeddings = []
for i in range(0, len(texts), batch):
text_batch = texts[i:(i+batch)]
# Embeddings model
emb_batch = emb_function(text_batch)
embeddings.append(emb_batch)
embeddings = np.vstack(embeddings)
return embeddings
embeddings = get_text_embedding(chunks)
print(embeddings.shape)
print(f\"Our text was embedded into {embeddings.shape[1]} dimensions\")<\/pre>\n\n\n\n\u53ef\u8996\u5316 Semantic Search<\/h3>\n\n\n\n
\u525b\u525b\u8f49\u63db\u6210 384 \u7dad\u5ea6\u7684\u5411\u91cf\uff0c\u6211\u5011\u4f7f\u7528 t-SNE \u7b97\u6cd5\uff0c\u964d\u5230\u4e8c\u7dad\uff0c\u8b93\u6211\u5011\u53ef\u4ee5\u5c31\u5716\u5f62\u505a\u89c0\u5bdf<\/p>\n\n\n\n
# Create a t-SNE model
tsne = TSNE(n_components=2, random_state=42)
embeddings_with_question = np.vstack([embeddings, emb_question])
embeddings_2d = tsne.fit_transform(embeddings_with_question)<\/pre>\n\n\n\ndef visualize_embeddings(embeddings_2d: np.ndarray,
question: Optional[bool] = False,
neighbors: Optional[np.ndarray] = None) -> None:
\"\"\"
Visualize 384-dimensional embeddings in 2D using t-SNE, label each data point with its index,
and optionally plot a question data point as a red dot with the label 'q'.
Args:
embeddings (numpy.array): An array of shape (num_samples, 384) containing the embeddings.
question (numpy.array, optional): An additional 384-dimensional embedding for the question.
Default is None.
\"\"\"
# Scatter plot the 2D embeddings and label each data point with its index
plt.figure(figsize=(10, 8))
num_samples = embeddings.shape[0]
if neighbors is not None:
for i, (x, y) in enumerate(embeddings_2d[:num_samples]):
if i in neighbors:
plt.scatter(x, y, color='purple', alpha=0.7)
plt.annotate(str(i), xy=(x, y), xytext=(5, 2), textcoords='offset points', color='black')
else:
plt.scatter(x, y, color='blue', alpha=0.7)
plt.annotate(str(i), xy=(x, y), xytext=(5, 2), textcoords='offset points', color='black')
else:
for i, (x, y) in enumerate(embeddings_2d[:num_samples]):
plt.scatter(x, y, color='blue', alpha=0.7)
plt.annotate(str(i), xy=(x, y), xytext=(5, 2), textcoords='offset points', color='black')
# Plot the question data point if provided
if question:
x, y = embeddings_2d[-1] # Last point corresponds to the question
plt.scatter(x, y, color='red', label='q')
plt.annotate('q', xy=(x, y), xytext=(5, 2), textcoords='offset points', color='black')
plt.title('t-SNE Visualization of 384-dimensional Embeddings')
plt.xlabel('Dimension 1')
plt.ylabel('Dimension 2')
plt.show()<\/pre>\n\n\n\nvisualize_embeddings(embeddings_2d[:-1])<\/pre>\n\n\n
\n![\"raw-image\"](\"https:\/\/d2a6d2ofes041u.cloudfront.net\/resize?compression=6&norotation=true&url=https%3A%2F%2Fimages.vocus.cc%2F9bfba657-8ff1-4b97-8906-a3a22db8b4c8.png&width=740&sign=W4fY79fI2-UBU9XDIsbFxgJmCgJaruWUs4WibrGf94Y\")
<\/figure>\n<\/div>\n\n\n<\/figure>\n<\/div>\n\n\n<\/figure>\n<\/div>\n\n\n\u6211\u5011\u8981\u900f\u904e\u6587\u5b57\u7247\u6bb5\u7684\u8ddd\u96e2\uff0c\u4f86\u627e\u5230\u53ef\u80fd\u5305\u542b\u554f\u984c\u7b54\u6848\u7684\u7247\u6bb5\uff0c\u9019\u88e1\u7684\u8ddd\u96e2\u9078\u7528\u6b50\u5e7e\u91cc\u5fb7\u8ddd\u96e2\u8a08\u7b97\u76f8\u4f3c\u5ea6\u8207\u63a5\u8fd1\u6027\uff0c\u63a1\u7528\u6700\u8fd1\u9130\u5c45<\/strong>(NearestNeighbors)\u7b97\u6cd5\uff0c\u9078\u64c7\u6700\u63a5\u8fd1\u7684 k \u500b\u6578\u64da\uff0ck \u5b9a\u70ba 5\u3002<\/p>\n\n\n\n
nn_2d = NearestNeighbors(n_neighbors=5, metric='cosine')
nn_2d.fit(embeddings_2d[:-1])<\/pre>\n\n\n\nneighbors = nn_2d.kneighbors(embeddings_2d[-1].reshape(1, -1), return_distance=False)
neighbors<\/pre>\n\n\n\n\u518d\u5c07\u5b83\u53ef\u8996\u5316<\/p>\n\n\n\n
visualize_embeddings(embeddings_2d, True, neighbors)<\/pre>\n\n\n
\n![\"raw-image\"](\"https:\/\/d2a6d2ofes041u.cloudfront.net\/resize?compression=6&norotation=true&url=https%3A%2F%2Fimages.vocus.cc%2F099547b8-c53b-4aee-86c5-b922fdf227fa.png&width=740&sign=lQz2X1u-553G6lUPgkuUmr9NrQotwaP1X3bfeRU8xCU\")
<\/figure>\n<\/div>\n\n\n<\/figure>\n<\/div>\n\n\n<\/figure>\n<\/div>\n\n\n\u7d2b\u8272\u7684\u9ede\u662f\u6bd4\u8f03\u9760\u8fd1\u7684\u3002<\/p>\n\n\n\n
\u4f7f\u7528 Semantic Search\u627e\u6700\u8fd1\u7684\u7247\u6bb5<\/h3>\n\n\n\n
question = '\u5982\u4f55\u67e5\u8a62\u6210\u4ea4\u660e\u7d30?'
emb_question = emb_function([question])<\/pre>\n\n\n\nneighbors = nn.kneighbors(emb_question, return_distance=False)
neighbors<\/pre>\n\n\n\n\u6e96\u5099 prompt<\/h3>\n\n\n\n
def build_prompt(question):
prompt = \"\"
prompt += \"[INST] <<SYS>>You are an AI assistant tasked with providing answers by summarizing related documents. You should follow these rules:\\n\"\\
\"1. Summarize the content from the provided documents, using the following format:\\n\"\\
\"Topic of the Document: Describe the topic of the document.\\n\"\\
\"Step by Step Instruction: Provide user question-specific instructions or information from the document.\\n\"\\
\"2. If no relevant information is found in the chat history, respond with \\\"I can't answer the question\\\".\\n\"\\
\"By adhering to these rules, you will help users find accurate and valuable information.\\n\"\\
\"<<\/SYS>>\\n\"
prompt += 'Search results:\\n'
for c in topn_chunks:
prompt += c + '\\n\\n'
prompt += f\"\\n\\n\\nQuery: {question}\\n\\nAnswer: [\/INST]\"
return prompt<\/pre>\n\n\n\nprompt = build_prompt(question)
print(prompt)<\/pre>\n\n\n\n![\"raw-image\"](\"https:\/\/d2a6d2ofes041u.cloudfront.net\/resize?compression=6&norotation=true&url=https%3A%2F%2Fimages.vocus.cc%2Fcf70011b-2c36-44d7-b793-28c04cdecad6.png&width=740&sign=2eFp0HliRUyClCHLHkSXaXLYJwG2Zslq8xLQmz_kLQ8\")
<\/figure>\n<\/div>\n\n\n<\/figure>\n<\/div>\n\n\n<\/figure>\n<\/div>\n\n\n\u6700\u5f8c\u4e86\uff01 inference<\/h3>\n\n\n\n
def send_to_watsonxai(prompts,
model_name=\"meta-llama\/llama-2-70b-chat\",
decoding_method=\"greedy\",
max_new_tokens=512,
min_new_tokens=0,
repetition_penalty=1.0
):
'''
helper function for sending prompts and params to Watsonx.ai
Args:
prompts:list list of text prompts
decoding:str Watsonx.ai parameter \"sample\" or \"greedy\"
max_new_tok:int Watsonx.ai parameter for max new tokens\/response returned
temperature:float Watsonx.ai parameter for temperature (range 0>2)
repetition_penalty:float Watsonx.ai parameter for repetition penalty (range 1.0 to 2.0)
Returns: None
prints response
'''
# Instantiate parameters for text generation
model_params = {
GenParams.DECODING_METHOD: decoding_method,
GenParams.MIN_NEW_TOKENS: min_new_tokens,
GenParams.MAX_NEW_TOKENS: max_new_tokens,
GenParams.RANDOM_SEED: 42,
GenParams.TEMPERATURE: 0,
GenParams.REPETITION_PENALTY: repetition_penalty,
}
# Instantiate a model proxy object to send your requests
model = Model(
model_id=model_name,
params=model_params,
credentials=creds,
project_id=project_id)
for prompt in prompts:
print(model.generate_text(prompt))<\/pre>\n\n\n\n# Example questions for the DQ2:
# \u53ef\u4ee5\u600e\u9ebc\u67e5\u8a62\u6210\u4ea4\u72c0\u6cc1?
# \u6709\u8fa6\u6cd5\u67e5\u8a62\u539f\u7269\u6599\u884c\u60c5\u55ce?
# \u6211\u60f3\u67e5\u8a62\u6307\u6578\u7684\u59d4\u8cb7\u8ddf\u59d4\u8ce3\u5206\u6790
# DQ2 \u53ef\u4ee5\u505a\u5230\u4ec0\u9ebc\u529f\u80fd?
question = \"\u6709\u8fa6\u6cd5\u67e5\u8a62\u539f\u7269\u6599\u884c\u60c5\u55ce?\"
prompt = build_prompt(question)
send_to_watsonxai(prompts=[prompt], min_new_tokens=1)<\/pre>\n\n\n\n![\"raw-image\"\/](\"https:\/\/d2a6d2ofes041u.cloudfront.net\/resize?compression=6&norotation=true&url=https%3A%2F%2Fimages.vocus.cc%2Fd4bcd74b-6e6f-4258-bd1c-3a137a4ed517.png&width=740&sign=F0EFDgjQ5Hc6Zd501lRt6XGgS97APy6lw1tk58RMjDU\")
<\/figure>\n<\/div>\n\n\n<\/figure>\n<\/div>\n\n\n<\/figure>\n<\/div>\n\n\n\u5c0f\u5fc3\u5f97<\/h2>\n\n\n\n
\u5728\u4e86\u89e3 RAG\uff08\u6aa2\u7d22\u589e\u5f37\u751f\u6210\uff09\u8207\u8a9e\u7fa9\u6aa2\u7d22\uff08Semantic Search\uff09\u6280\u8853\u5f8c\uff0c\u6211\u767c\u73fe\u96d6\u7136\u5b83\u5011\u5728\u6280\u8853\u5c64\u9762\u4e0a\u6bd4\u4e4b\u524d\u6211\u5206\u4eab\u7684\u5167\u5bb9\u4f86\u5f97\u8907\u96dc\uff0c\u4f46\u5be6\u969b\u4e0a\u5176\u6838\u5fc3\u6982\u5ff5\u975e\u5e38\u660e\u78ba\u4e14\u76f4\u89c0\u3002\u9019\u4e9b\u6280\u8853\u7684\u51fa\u73fe\uff0c\u4e3b\u8981\u662f\u70ba\u4e86\u89e3\u6c7a\u7576\u524d\u5927\u578b\u8a9e\u8a00\u6a21\u578b\uff08LLM\uff09\u5728\u8655\u7406\u5373\u6642\u66f4\u65b0\u8cc7\u8a0a\u6216\u7279\u5b9a\u9818\u57df\u6df1\u5ea6\u77e5\u8b58\u65b9\u9762\u7684\u5c40\u9650\u6027\u3002RAG \u548c\u8a9e\u7fa9\u6aa2\u7d22\u9019\u4e9b\u201c\u5de7\u5999\u7684\u65b9\u6cd5\u201d\u5be6\u969b\u4e0a\u4ee3\u8868\u4e86\u5728\u73fe\u6709\u6280\u8853\u689d\u4ef6\u4e0b\u7684\u5275\u65b0\u8207\u7a81\u7834\u3002\u6211\u8a8d\u70ba\uff0c\u96a8\u8457\u6280\u8853\u7684\u4e0d\u65b7\u767c\u5c55\uff0c\u9019\u4e9b\u9650\u5236\u5f88\u53ef\u80fd\u6703\u5f88\u5feb\u88ab\u514b\u670d\u3002\u8207\u6b64\u540c\u6642\uff0c\u9019\u4e9b\u65b9\u6cd5\u5728\u7bc0\u7d04\u8a08\u7b97\u8cc7\u6e90\u548c\u63d0\u9ad8\u56de\u7b54\u7cbe\u78ba\u5ea6\u65b9\u9762\u7684\u6f5b\u529b\u4e5f\u4ee4\u6211\u89ba\u5f97\u5b83\u5011\u975e\u5e38\u503c\u5f97\u5617\u8a66\u548c\u6df1\u5165\u63a2\u7a76\uff0cRAG \u548c\u8a9e\u7fa9\u6aa2\u7d22\u4e0d\u50c5\u662f\u7576\u524d\u554f\u984c\u7684\u6709\u6548\u89e3\u6c7a\u65b9\u6848\uff0c\u4e5f\u53ef\u80fd\u70ba\u672a\u4f86\u8a9e\u8a00\u6a21\u578b\u7684\u767c\u5c55\u63d0\u4f9b\u65b0\u7684\u601d\u8def\u548c\u9748\u611f\uff0c\u611f\u8b1d\u770b\u5230\u9019\u88e1\u7684\u4f60\uff0c\u6211\u5011\u4e0b\u6b21\u898b\uff01<\/p>\n\n\n\n
\u53c3\u8003<\/h2>\n\n\n\n
- \n
- RAG \u6aa2\u7d22\u589e\u5f37\u751f\u6210\u2014 \u8b93\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u66f4\u8070\u660e\u7684\u79d8\u5bc6\u6b66\u5668. \u770b\u5b8c\u6587\u7ae0\u5f8c\u6b61\u8fce\u6309\u9f13\u52f5\uff0c\u8a02\u95b1\uff0c\u4e26\u5206\u4eab\u7d66\u6240\u6709\u60f3\u77e5\u9053\u6b64\u985e\u77e5\u8b58\u7684\u6240\u6709\u4eba\uff01 | by SimonLiu | Nov, 2023 | InfuseAI<\/li>\n\n\n\n
- Universal Sentence Encoder<\/li>\n\n\n\n
- Lab watsonx_pdf_rag<\/li>\n\n\n\n
- \u5168\u7aef LLM \u61c9\u7528\u958b\u767c-Day29-Langchain \u7684 chain<\/li>\n<\/ul>\n\n\n\n
- \u63d0\u4f9b\u66f4\u6e96\u78ba\u548c\u76f8\u95dc\u7684\u56de\u7b54\uff1a<\/strong> \u901a\u904e\u6aa2\u7d22\u548c\u6574\u5408\u8207\u7528\u6236\u67e5\u8a62\u76f8\u95dc\u7684\u5916\u90e8\u4fe1\u606f\uff0cRAG \u80fd\u5920\u4f7f\u6a21\u578b\u63d0\u4f9b\u66f4\u6e96\u78ba\u3001\u66f4\u5177\u76f8\u95dc\u6027\u7684\u7b54\u6848\u3002\u9019\u5c0d\u65bc\u9700\u8981\u5c08\u696d\u77e5\u8b58\u6216\u7279\u5b9a\u9818\u57df\u4fe1\u606f\u7684\u67e5\u8a62\u7279\u5225\u6709\u7528\uff0c\u9019\u4e9b\u5c08\u696d\u77e5\u8b58\u53ef\u80fd\u4f86\u81ea\u67d0\u4e9b\u9577\u7bc7\u7684\u6587\u4ef6\u4e2d\uff0c\u90a3\u5c31\u5f88\u9069\u5408\u900f\u904e RAG \u53bb\u6aa2\u7d22\u8207\u751f\u6210\uff0c\u4e14\u751f\u6210\u7684\u5167\u5bb9\u4e5f\u6703\u6839\u64da\u6b64\u9577\u6587\uff0c\u5c0e\u51fa\u7cbe\u78ba\u7684\u56de\u7b54\u3002<\/li>\n\n\n\n
- \u8655\u7406\u5373\u6642\u6216\u6700\u65b0\u4fe1\u606f\uff1a<\/strong> \u50b3\u7d71\u7684\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u5728\u7279\u5b9a\u6642\u9593\u9ede\u7684\u6578\u64da\u5feb\u7167\u4e0a\u9032\u884c\u8a13\u7df4\uff0c\u56e0\u6b64\u5b83\u5011\u50c5\u80fd\u56de\u7b54\u8a13\u7df4\u6578\u64da\u4e2d\u5305\u542b\u7684\u4fe1\u606f\u3002\u4f8b\u5982\uff0c\u554f\u4e00\u500b LLM \u4eca\u5929\u7684\u5929\u6c23\u5982\u4f55\uff0c\u5b83\u53ef\u80fd\u7121\u6cd5\u56de\u7b54\uff0c\u56e0\u70ba\u4eca\u5929\u7684\u5929\u6c23\u4fe1\u606f\u4e0d\u5305\u542b\u5728\u5b83\u7684\u8a13\u7df4\u6578\u64da\u4e2d\u3002RAG \u900f\u904e\u63a2\u7d22\u5373\u6642\u5929\u6c23\u7684 API \u6216\u76f8\u95dc\u4fe1\u606f\u4f86\u7372\u5f97\u8cc7\u8a0a\uff0c\u4e26\u5728\u95b1\u8b80\u9019\u4e9b API \u7684\u7d50\u679c\u5f8c\uff0c\u751f\u6210\u9700\u8981\u7684\u7b54\u6848\uff0c\u4ee5\u5f4c\u88dc\u8655\u7406\u5373\u6642\u554f\u984c\u7684\u9650\u5236\u3002<\/li>\n\n\n\n