Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -17,14 +17,6 @@ from llama_cpp_agent.prompt_templates import web_search_system_prompt, research_
|
|
17 |
from style import css, PLACEHOLDER
|
18 |
from utils import CitingSources
|
19 |
|
20 |
-
model_selected = "Mistral-7B-Instruct-v0.3-Q6_K.gguf"
|
21 |
-
examples = [
|
22 |
-
["latest news about Yann LeCun"],
|
23 |
-
["Latest news site:github.blog"],
|
24 |
-
["Where I can find best hotel in Galapagos, Ecuador intitle:hotel"],
|
25 |
-
["filetype:pdf intitle:python"]
|
26 |
-
]
|
27 |
-
|
28 |
hf_hub_download(
|
29 |
repo_id="bartowski/Mistral-7B-Instruct-v0.3-GGUF",
|
30 |
filename="Mistral-7B-Instruct-v0.3-Q6_K.gguf",
|
@@ -35,6 +27,18 @@ hf_hub_download(
|
|
35 |
filename="Meta-Llama-3-8B-Instruct-Q6_K.gguf",
|
36 |
local_dir="./models"
|
37 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
|
39 |
css = """
|
40 |
.message-row {
|
@@ -98,19 +102,20 @@ PLACEHOLDER = """
|
|
98 |
def get_context_by_model(model_name):
|
99 |
model_context_limits = {
|
100 |
"Mistral-7B-Instruct-v0.3-Q6_K.gguf": 32768,
|
|
|
101 |
"Meta-Llama-3-8B-Instruct-Q6_K.gguf": 8192
|
102 |
}
|
103 |
return model_context_limits.get(model_name, None)
|
104 |
|
105 |
def get_messages_formatter_type(model_name):
|
106 |
-
|
107 |
-
if
|
108 |
return MessagesFormatterType.LLAMA_3
|
109 |
-
elif
|
110 |
return MessagesFormatterType.MISTRAL
|
111 |
-
elif
|
112 |
return MessagesFormatterType.CHATML
|
113 |
-
elif "
|
114 |
return MessagesFormatterType.PHI_3
|
115 |
else:
|
116 |
return MessagesFormatterType.CHATML
|
@@ -236,6 +241,7 @@ demo = gr.ChatInterface(
|
|
236 |
additional_inputs=[
|
237 |
gr.Dropdown([
|
238 |
'Mistral-7B-Instruct-v0.3-Q6_K.gguf',
|
|
|
239 |
'Meta-Llama-3-8B-Instruct-Q6_K.gguf'
|
240 |
],
|
241 |
value="Mistral-7B-Instruct-v0.3-Q6_K.gguf",
|
|
|
17 |
from style import css, PLACEHOLDER
|
18 |
from utils import CitingSources
|
19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
hf_hub_download(
|
21 |
repo_id="bartowski/Mistral-7B-Instruct-v0.3-GGUF",
|
22 |
filename="Mistral-7B-Instruct-v0.3-Q6_K.gguf",
|
|
|
27 |
filename="Meta-Llama-3-8B-Instruct-Q6_K.gguf",
|
28 |
local_dir="./models"
|
29 |
)
|
30 |
+
hf_hub_download(
|
31 |
+
repo_id="TheBloke/Mixtral-8x7B-Instruct-v0.1-GGUF",
|
32 |
+
filename="mixtral-8x7b-instruct-v0.1.Q5_K_M.gguf",
|
33 |
+
local_dir="./models"
|
34 |
+
)
|
35 |
+
|
36 |
+
examples = [
|
37 |
+
["latest news about Yann LeCun"],
|
38 |
+
["Latest news site:github.blog"],
|
39 |
+
["Where I can find best hotel in Galapagos, Ecuador intitle:hotel"],
|
40 |
+
["filetype:pdf intitle:python"]
|
41 |
+
]
|
42 |
|
43 |
css = """
|
44 |
.message-row {
|
|
|
102 |
def get_context_by_model(model_name):
|
103 |
model_context_limits = {
|
104 |
"Mistral-7B-Instruct-v0.3-Q6_K.gguf": 32768,
|
105 |
+
"mixtral-8x7b-instruct-v0.1.Q5_K_M.gguf": 32768,
|
106 |
"Meta-Llama-3-8B-Instruct-Q6_K.gguf": 8192
|
107 |
}
|
108 |
return model_context_limits.get(model_name, None)
|
109 |
|
110 |
def get_messages_formatter_type(model_name):
|
111 |
+
model_name = model_name.lower()
|
112 |
+
if any(keyword in model_name for keyword in ["meta", "aya"]):
|
113 |
return MessagesFormatterType.LLAMA_3
|
114 |
+
elif any(keyword in model_name for keyword in ["mistral", "mixtral"]):
|
115 |
return MessagesFormatterType.MISTRAL
|
116 |
+
elif any(keyword in model_name for keyword in ["einstein", "dolphin"]):
|
117 |
return MessagesFormatterType.CHATML
|
118 |
+
elif "phi" in model_name:
|
119 |
return MessagesFormatterType.PHI_3
|
120 |
else:
|
121 |
return MessagesFormatterType.CHATML
|
|
|
241 |
additional_inputs=[
|
242 |
gr.Dropdown([
|
243 |
'Mistral-7B-Instruct-v0.3-Q6_K.gguf',
|
244 |
+
'mixtral-8x7b-instruct-v0.1.Q5_K_M.gguf',
|
245 |
'Meta-Llama-3-8B-Instruct-Q6_K.gguf'
|
246 |
],
|
247 |
value="Mistral-7B-Instruct-v0.3-Q6_K.gguf",
|