Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -21,6 +21,18 @@ def transcribe(audio):
|
|
21 |
|
22 |
HF_TOKEN = os.environ.get("HF_TOKEN", None)
|
23 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
24 |
def randomize_seed_fn(seed: int) -> int:
|
25 |
seed = random.randint(0, 999999)
|
26 |
return seed
|
@@ -33,18 +45,17 @@ Respond in a normal, conversational manner while being friendly and helpful.
|
|
33 |
[USER]
|
34 |
"""
|
35 |
|
36 |
-
def models(text, seed=42):
|
37 |
|
38 |
seed = int(randomize_seed_fn(seed))
|
39 |
generator = torch.Generator().manual_seed(seed)
|
40 |
|
41 |
-
client =
|
42 |
|
43 |
generate_kwargs = dict(
|
44 |
max_new_tokens=300,
|
45 |
seed=seed
|
46 |
-
)
|
47 |
-
|
48 |
formatted_prompt = system_instructions1 + text + "[JARVIS]"
|
49 |
stream = client.text_generation(
|
50 |
formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
|
@@ -52,7 +63,6 @@ def models(text, seed=42):
|
|
52 |
for response in stream:
|
53 |
if not response.token.text == "</s>":
|
54 |
output += response.token.text
|
55 |
-
|
56 |
return output
|
57 |
|
58 |
async def respond(audio, model, seed):
|
@@ -72,6 +82,14 @@ DESCRIPTION = """ # <center><b>JARVIS⚡</b></center>
|
|
72 |
with gr.Blocks(css="style.css") as demo:
|
73 |
gr.Markdown(DESCRIPTION)
|
74 |
with gr.Row():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
seed = gr.Slider(
|
76 |
label="Seed",
|
77 |
minimum=0,
|
@@ -89,8 +107,8 @@ with gr.Blocks(css="style.css") as demo:
|
|
89 |
batch=True,
|
90 |
max_batch_size=10,
|
91 |
fn=respond,
|
92 |
-
inputs=[input, seed],
|
93 |
outputs=[output], live=True)
|
94 |
-
|
95 |
if __name__ == "__main__":
|
96 |
demo.queue(max_size=200).launch()
|
|
|
21 |
|
22 |
HF_TOKEN = os.environ.get("HF_TOKEN", None)
|
23 |
|
24 |
+
def client_fn(model):
|
25 |
+
if "Mixtral" in model:
|
26 |
+
return InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
|
27 |
+
elif "Llama" in model:
|
28 |
+
return InferenceClient("meta-llama/Meta-Llama-3-8B-Instruct")
|
29 |
+
elif "Mistral" in model:
|
30 |
+
return InferenceClient("mistralai/Mistral-7B-Instruct-v0.2")
|
31 |
+
elif "Phi" in model:
|
32 |
+
return InferenceClient("microsoft/Phi-3-mini-4k-instruct")
|
33 |
+
else:
|
34 |
+
return InferenceClient("microsoft/Phi-3-mini-4k-instruct")
|
35 |
+
|
36 |
def randomize_seed_fn(seed: int) -> int:
|
37 |
seed = random.randint(0, 999999)
|
38 |
return seed
|
|
|
45 |
[USER]
|
46 |
"""
|
47 |
|
48 |
+
def models(text, model="Mixtral 8x7B", seed=42):
|
49 |
|
50 |
seed = int(randomize_seed_fn(seed))
|
51 |
generator = torch.Generator().manual_seed(seed)
|
52 |
|
53 |
+
client = client_fn(model)
|
54 |
|
55 |
generate_kwargs = dict(
|
56 |
max_new_tokens=300,
|
57 |
seed=seed
|
58 |
+
)
|
|
|
59 |
formatted_prompt = system_instructions1 + text + "[JARVIS]"
|
60 |
stream = client.text_generation(
|
61 |
formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
|
|
|
63 |
for response in stream:
|
64 |
if not response.token.text == "</s>":
|
65 |
output += response.token.text
|
|
|
66 |
return output
|
67 |
|
68 |
async def respond(audio, model, seed):
|
|
|
82 |
with gr.Blocks(css="style.css") as demo:
|
83 |
gr.Markdown(DESCRIPTION)
|
84 |
with gr.Row():
|
85 |
+
select = gr.Dropdown([ 'Mixtral 8x7B',
|
86 |
+
'Llama 3 8B',
|
87 |
+
'Mistral 7B v0.3',
|
88 |
+
'Phi 3 mini',
|
89 |
+
],
|
90 |
+
value="Mistral 7B v0.3",
|
91 |
+
label="Model"
|
92 |
+
)
|
93 |
seed = gr.Slider(
|
94 |
label="Seed",
|
95 |
minimum=0,
|
|
|
107 |
batch=True,
|
108 |
max_batch_size=10,
|
109 |
fn=respond,
|
110 |
+
inputs=[input, select, seed],
|
111 |
outputs=[output], live=True)
|
112 |
+
|
113 |
if __name__ == "__main__":
|
114 |
demo.queue(max_size=200).launch()
|