Update app.py
Browse files
app.py
CHANGED
|
@@ -1,6 +1,7 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
import time
|
| 3 |
from transformers import AutoModelForCausalLM # Please ensure this import is correct
|
|
|
|
| 4 |
|
| 5 |
PROMPT_TEMPLATE = (
|
| 6 |
"<s>" "[INST]"
|
|
@@ -18,7 +19,7 @@ PROMPT_TEMPLATE = (
|
|
| 18 |
)
|
| 19 |
|
| 20 |
def load_llm():
|
| 21 |
-
llm = AutoModelForCausalLM.from_pretrained("
|
| 22 |
model_type='llama',
|
| 23 |
max_new_tokens = 1096,
|
| 24 |
repetition_penalty = 1.13,
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
import time
|
| 3 |
from transformers import AutoModelForCausalLM # Please ensure this import is correct
|
| 4 |
+
from download_model import download_model
|
| 5 |
|
| 6 |
PROMPT_TEMPLATE = (
|
| 7 |
"<s>" "[INST]"
|
|
|
|
| 19 |
)
|
| 20 |
|
| 21 |
def load_llm():
|
| 22 |
+
llm = AutoModelForCausalLM.from_pretrained("Llama-2-7b-chat-q8-gguf",
|
| 23 |
model_type='llama',
|
| 24 |
max_new_tokens = 1096,
|
| 25 |
repetition_penalty = 1.13,
|