Spaces:
Sleeping
Sleeping
emar
commited on
Commit
•
d829f59
1
Parent(s):
b7463ad
return to bge
Browse files- app.py +17 -7
- storage/default__vector_store.json +2 -2
- storage/docstore.json +1 -1
- storage/index_store.json +1 -1
app.py
CHANGED
@@ -15,8 +15,7 @@ PERSIST_DIR = './storage'
|
|
15 |
# Configure the settings
|
16 |
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
17 |
|
18 |
-
Settings.embed_model =
|
19 |
-
# Settings.embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-base-en-v1.5", device="cpu")
|
20 |
|
21 |
Settings.llm = HuggingFaceLLM(
|
22 |
model_name="meta-llama/Meta-Llama-3-8B-Instruct",
|
@@ -39,17 +38,28 @@ query_engine = index.as_query_engine(streaming=True, similarity_top_k=1, node_po
|
|
39 |
|
40 |
|
41 |
@spaces.GPU
|
42 |
-
def chatbot_response(message,
|
|
|
|
|
|
|
43 |
response = query_engine.query(message)
|
44 |
return str(response)
|
45 |
|
46 |
-
iface = gr.
|
47 |
fn=chatbot_response,
|
48 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
description="Github page for use case, general information, local installs, etc: https://github.com/emarron/UESP-lore",
|
50 |
-
examples=["Who is Zaraphus?", "What is the relation between dragonbreak and chim?", "What is the Lunar Lorkhan?"],
|
51 |
-
cache_examples=True,
|
52 |
)
|
53 |
|
|
|
|
|
54 |
if __name__ == "__main__":
|
55 |
iface.launch()
|
|
|
15 |
# Configure the settings
|
16 |
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
17 |
|
18 |
+
Settings.embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-base-en-v1.5", device="cpu")
|
|
|
19 |
|
20 |
Settings.llm = HuggingFaceLLM(
|
21 |
model_name="meta-llama/Meta-Llama-3-8B-Instruct",
|
|
|
38 |
|
39 |
|
40 |
@spaces.GPU
|
41 |
+
def chatbot_response(message, context_window, max_new_tokens, temperature, top_k, top_p):
|
42 |
+
Settings.llm.context_window = context_window
|
43 |
+
Settings.llm.max_new_tokens = max_new_tokens
|
44 |
+
Settings.llm.generate_kwargs = {"temperature": temperature, "top_k": top_k, "top_p": top_p, "do sample": True}
|
45 |
response = query_engine.query(message)
|
46 |
return str(response)
|
47 |
|
48 |
+
iface = gr.Interface(
|
49 |
fn=chatbot_response,
|
50 |
+
inputs=[
|
51 |
+
gr.Slider(minimum=512, maximum=4096, step=256, value=2048, label="Context Window"),
|
52 |
+
gr.Slider(minimum=32, maximum=512, step=32, value=256, label="Max New Tokens"),
|
53 |
+
gr.Slider(minimum=0.1, maximum=1.0, step=0.1, value=0.7, label="Temperature: Lower number is by the book, higher number is make stuff up"),
|
54 |
+
gr.Slider(minimum=1, maximum=100, step=1, value=50, label="Top K"),
|
55 |
+
gr.Slider(minimum=0.5, maximum=1.0, step=0.05, value=0.95, label="Top P"),
|
56 |
+
],
|
57 |
+
outputs=gr.Textbox(label="Response"),
|
58 |
+
title="UESP Lore Chatbot: Running on top of Meta-Llama-3-8B-Instruct + BGE_LARGE. This is inferior to the downloadable ones.",
|
59 |
description="Github page for use case, general information, local installs, etc: https://github.com/emarron/UESP-lore",
|
|
|
|
|
60 |
)
|
61 |
|
62 |
+
|
63 |
+
|
64 |
if __name__ == "__main__":
|
65 |
iface.launch()
|
storage/default__vector_store.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9d5b72dceac75e8648eeafd9a9e1b6d62aa24ef560b3fb1116e0e43c65d5ccb4
|
3 |
+
size 197975710
|
storage/docstore.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 44043711
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f9793071f910c4d3c66e206d6cf0930d9a869c243f1fad99e19b25d3612df161
|
3 |
size 44043711
|
storage/index_store.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 959947
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c31d504b2c3a15e815dcfe5af9215cb8bbd7e5d5530cf0cd7d0dd8511413477a
|
3 |
size 959947
|