Spaces:
Sleeping
Sleeping
Pclanglais
commited on
Commit
•
208476f
1
Parent(s):
459a15e
Update app.py
Browse files
app.py
CHANGED
@@ -14,6 +14,8 @@ from chromadb.utils import embedding_functions
|
|
14 |
from FlagEmbedding import BGEM3FlagModel
|
15 |
from sklearn.metrics.pairwise import cosine_similarity
|
16 |
|
|
|
|
|
17 |
model = BGEM3FlagModel('BAAI/bge-m3',
|
18 |
use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation
|
19 |
|
@@ -22,16 +24,16 @@ embeddings_data = pd.read_json("embeddings_tchap.json")
|
|
22 |
embeddings_text = embeddings_data["text_with_context"].tolist()
|
23 |
|
24 |
# Define the device
|
25 |
-
#device = "cuda" if torch.cuda.is_available() else "cpu"
|
26 |
-
#Define variables
|
27 |
temperature=0.2
|
28 |
max_new_tokens=1000
|
29 |
top_p=0.92
|
30 |
repetition_penalty=1.7
|
31 |
|
32 |
-
|
33 |
|
34 |
-
|
|
|
|
|
35 |
|
36 |
system_prompt = "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nTu es Albert, l'agent conversationnel des services publics qui peut décrire des documents de référence ou aider à des tâches de rédaction<|eot_id|>"
|
37 |
|
@@ -78,7 +80,7 @@ def predict(message, history):
|
|
78 |
|
79 |
messages = system_prompt + messages
|
80 |
|
81 |
-
|
82 |
streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
|
83 |
generate_kwargs = dict(
|
84 |
model_inputs,
|
@@ -98,7 +100,7 @@ def predict(message, history):
|
|
98 |
for new_token in streamer:
|
99 |
if new_token != '<':
|
100 |
partial_message += new_token
|
101 |
-
yield partial_message
|
102 |
return messages
|
103 |
|
104 |
# Define the Gradio interface
|
|
|
14 |
from FlagEmbedding import BGEM3FlagModel
|
15 |
from sklearn.metrics.pairwise import cosine_similarity
|
16 |
|
17 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
18 |
+
|
19 |
model = BGEM3FlagModel('BAAI/bge-m3',
|
20 |
use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation
|
21 |
|
|
|
24 |
embeddings_text = embeddings_data["text_with_context"].tolist()
|
25 |
|
26 |
# Define the device
|
|
|
|
|
27 |
temperature=0.2
|
28 |
max_new_tokens=1000
|
29 |
top_p=0.92
|
30 |
repetition_penalty=1.7
|
31 |
|
32 |
+
model_name = "Pclanglais/Tchap"
|
33 |
|
34 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
35 |
+
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16)
|
36 |
+
model = model.to('cuda:0')
|
37 |
|
38 |
system_prompt = "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nTu es Albert, l'agent conversationnel des services publics qui peut décrire des documents de référence ou aider à des tâches de rédaction<|eot_id|>"
|
39 |
|
|
|
80 |
|
81 |
messages = system_prompt + messages
|
82 |
|
83 |
+
model_inputs = tokenizer([messages], return_tensors="pt").to("cuda")
|
84 |
streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
|
85 |
generate_kwargs = dict(
|
86 |
model_inputs,
|
|
|
100 |
for new_token in streamer:
|
101 |
if new_token != '<':
|
102 |
partial_message += new_token
|
103 |
+
yield partial_message
|
104 |
return messages
|
105 |
|
106 |
# Define the Gradio interface
|