Spaces:
Running
on
Zero
Running
on
Zero
alfredplpl
commited on
Commit
•
e32c4ee
1
Parent(s):
a915b3f
Update app.py
Browse files
app.py
CHANGED
@@ -12,7 +12,7 @@ from threading import Thread
|
|
12 |
DESCRIPTION = '''
|
13 |
<div>
|
14 |
<h1 style="text-align: center;">非公式Llama-3.1-Swallow-8B-Instruct-v0.1</h1>
|
15 |
-
<p>tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.1の非公式デモだよ。 <a href="https://huggingface.co/tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.1"><b>tokyotech-llm/Llama-3-Swallow-8B-Instruct-v0.1</b></a>.</p>
|
16 |
</div>
|
17 |
'''
|
18 |
|
@@ -25,7 +25,7 @@ Built with Meta Llama 3.1
|
|
25 |
|
26 |
PLACEHOLDER = """
|
27 |
<div style="padding: 30px; text-align: center; display: flex; flex-direction: column; align-items: center;">
|
28 |
-
<h1 style="font-size: 28px; margin-bottom: 2px; opacity: 0.55;">Meta llama3</h1>
|
29 |
<p style="font-size: 18px; margin-bottom: 2px; opacity: 0.65;">なんでもきいてね</p>
|
30 |
</div>
|
31 |
"""
|
@@ -48,10 +48,6 @@ h1 {
|
|
48 |
# Load the tokenizer and model
|
49 |
tokenizer = AutoTokenizer.from_pretrained("tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.1")
|
50 |
model = AutoModelForCausalLM.from_pretrained("tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.1", torch_dtype=torch.bfloat16, device_map="auto")
|
51 |
-
terminators = [
|
52 |
-
tokenizer.eos_token_id,
|
53 |
-
tokenizer.convert_tokens_to_ids("<|eot_id|>")
|
54 |
-
]
|
55 |
|
56 |
@spaces.GPU()
|
57 |
def chat_llama3_8b(message: str,
|
@@ -87,7 +83,6 @@ def chat_llama3_8b(message: str,
|
|
87 |
temperature=temperature,
|
88 |
top_p=0.9,
|
89 |
repetition_penalty=1.1,
|
90 |
-
eos_token_id=terminators,
|
91 |
)
|
92 |
# This will enforce greedy generation (do_sample=False) when the temperature is passed 0, avoiding the crash.
|
93 |
if temperature == 0:
|
@@ -119,7 +114,7 @@ with gr.Blocks(fill_height=True, css=css) as demo:
|
|
119 |
gr.Slider(minimum=0,
|
120 |
maximum=1,
|
121 |
step=0.1,
|
122 |
-
value=0.
|
123 |
label="Temperature",
|
124 |
render=False),
|
125 |
gr.Slider(minimum=128,
|
|
|
12 |
DESCRIPTION = '''
|
13 |
<div>
|
14 |
<h1 style="text-align: center;">非公式Llama-3.1-Swallow-8B-Instruct-v0.1</h1>
|
15 |
+
<p>tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.1の非公式デモだよ。 <a href="https://huggingface.co/tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.1"><b>tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.1</b></a>.</p>
|
16 |
</div>
|
17 |
'''
|
18 |
|
|
|
25 |
|
26 |
PLACEHOLDER = """
|
27 |
<div style="padding: 30px; text-align: center; display: flex; flex-direction: column; align-items: center;">
|
28 |
+
<h1 style="font-size: 28px; margin-bottom: 2px; opacity: 0.55;">Meta llama3.1</h1>
|
29 |
<p style="font-size: 18px; margin-bottom: 2px; opacity: 0.65;">なんでもきいてね</p>
|
30 |
</div>
|
31 |
"""
|
|
|
48 |
# Load the tokenizer and model
|
49 |
tokenizer = AutoTokenizer.from_pretrained("tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.1")
|
50 |
model = AutoModelForCausalLM.from_pretrained("tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.1", torch_dtype=torch.bfloat16, device_map="auto")
|
|
|
|
|
|
|
|
|
51 |
|
52 |
@spaces.GPU()
|
53 |
def chat_llama3_8b(message: str,
|
|
|
83 |
temperature=temperature,
|
84 |
top_p=0.9,
|
85 |
repetition_penalty=1.1,
|
|
|
86 |
)
|
87 |
# This will enforce greedy generation (do_sample=False) when the temperature is passed 0, avoiding the crash.
|
88 |
if temperature == 0:
|
|
|
114 |
gr.Slider(minimum=0,
|
115 |
maximum=1,
|
116 |
step=0.1,
|
117 |
+
value=0.6,
|
118 |
label="Temperature",
|
119 |
render=False),
|
120 |
gr.Slider(minimum=128,
|