Spaces:
Sleeping
Sleeping
Add reranking + RAGAS evaluation
Browse files- requirements-eval.txt +22 -0
requirements-eval.txt
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Offline evaluation + notebook only — NOT needed to run the app.
|
| 2 |
+
# The app reads the pre-computed eval_results.json.
|
| 3 |
+
#
|
| 4 |
+
# Install:
|
| 5 |
+
# pip install -r requirements.txt
|
| 6 |
+
# pip install --no-deps ragas # --no-deps: skip scikit-network, a
|
| 7 |
+
# # knowledge-graph dep we don't use that
|
| 8 |
+
# # has no Python 3.14 wheel
|
| 9 |
+
# pip install -r requirements-eval.txt
|
| 10 |
+
# python evaluate.py
|
| 11 |
+
#
|
| 12 |
+
# Note: evaluate.py shims langchain_community.chat_models.vertexai (removed in
|
| 13 |
+
# langchain-community 0.4.x) so RAGAS 0.4.x imports cleanly on the LangChain 1.x
|
| 14 |
+
# stack — no downgrade of the app's dependencies required.
|
| 15 |
+
|
| 16 |
+
instructor>=1.0.0
|
| 17 |
+
datasets>=4.0.0
|
| 18 |
+
langchain-openai>=0.2.0
|
| 19 |
+
tiktoken>=0.7.0
|
| 20 |
+
pysbd>=0.3.4
|
| 21 |
+
matplotlib>=3.7.0
|
| 22 |
+
jupyter>=1.0.0
|