Spaces:
Sleeping
Sleeping
Dylan
commited on
Commit
·
350f8a0
1
Parent(s):
0c35e90
calling gemma
Browse files
agents.py
CHANGED
@@ -31,6 +31,9 @@ def build_graph():
|
|
31 |
workflow = StateGraph(State)
|
32 |
|
33 |
# Add nodes
|
|
|
|
|
|
|
34 |
workflow.add_node("caption_image", caption_image)
|
35 |
workflow.add_node("describe_with_voice", describe_with_voice)
|
36 |
|
@@ -55,16 +58,21 @@ model = Gemma3ForConditionalGeneration.from_pretrained(
|
|
55 |
)
|
56 |
|
57 |
|
58 |
-
def
|
59 |
-
|
|
|
|
|
60 |
return state
|
61 |
|
62 |
|
63 |
-
def
|
64 |
-
|
|
|
|
|
|
|
65 |
|
66 |
|
67 |
-
def
|
68 |
caption = state["caption"]
|
69 |
voice = state["voice"]
|
70 |
|
@@ -75,7 +83,10 @@ def describe_with_voice2(state: State) -> State:
|
|
75 |
"sarcastic teenager": "You are a sarcastic and disinterested teenager.",
|
76 |
}
|
77 |
messages = [
|
78 |
-
{
|
|
|
|
|
|
|
79 |
{
|
80 |
"role": "user",
|
81 |
"content": [
|
@@ -93,7 +104,7 @@ def describe_with_voice2(state: State) -> State:
|
|
93 |
input_len = inputs["input_ids"].shape[-1]
|
94 |
|
95 |
with torch.inference_mode():
|
96 |
-
generation = model.generate(**inputs, max_new_tokens=
|
97 |
generation = generation[0][input_len:]
|
98 |
|
99 |
description = processor.decode(generation, skip_special_tokens=True)
|
@@ -103,7 +114,7 @@ def describe_with_voice2(state: State) -> State:
|
|
103 |
return state
|
104 |
|
105 |
|
106 |
-
def
|
107 |
# image is PIL
|
108 |
image = state["image"]
|
109 |
|
@@ -136,7 +147,7 @@ def caption_image2(state: State) -> State:
|
|
136 |
input_len = inputs["input_ids"].shape[-1]
|
137 |
|
138 |
with torch.inference_mode():
|
139 |
-
generation = model.generate(**inputs, max_new_tokens=
|
140 |
generation = generation[0][input_len:]
|
141 |
|
142 |
caption = processor.decode(generation, skip_special_tokens=True)
|
|
|
31 |
workflow = StateGraph(State)
|
32 |
|
33 |
# Add nodes
|
34 |
+
# workflow.add_node("caption_image", caption_image_dummy)
|
35 |
+
# workflow.add_node("describe_with_voice", describe_with_voice_dummy)
|
36 |
+
|
37 |
workflow.add_node("caption_image", caption_image)
|
38 |
workflow.add_node("describe_with_voice", describe_with_voice)
|
39 |
|
|
|
58 |
)
|
59 |
|
60 |
|
61 |
+
def describe_with_voice_dummy(state: State) -> State:
|
62 |
+
print("Describe")
|
63 |
+
voice = state["voice"]
|
64 |
+
state["description"] = f"Dummy description from {voice}"
|
65 |
return state
|
66 |
|
67 |
|
68 |
+
def caption_image_dummy(state: State) -> State:
|
69 |
+
print("Caption")
|
70 |
+
voice = state["voice"]
|
71 |
+
state["caption"] = f"Dummy caption from {voice}"
|
72 |
+
return state
|
73 |
|
74 |
|
75 |
+
def describe_with_voice(state: State) -> State:
|
76 |
caption = state["caption"]
|
77 |
voice = state["voice"]
|
78 |
|
|
|
83 |
"sarcastic teenager": "You are a sarcastic and disinterested teenager.",
|
84 |
}
|
85 |
messages = [
|
86 |
+
{
|
87 |
+
"role": "system",
|
88 |
+
"content": [{"type": "text", "text": voice_prompts.get(voice)}],
|
89 |
+
},
|
90 |
{
|
91 |
"role": "user",
|
92 |
"content": [
|
|
|
104 |
input_len = inputs["input_ids"].shape[-1]
|
105 |
|
106 |
with torch.inference_mode():
|
107 |
+
generation = model.generate(**inputs, max_new_tokens=1000, do_sample=False)
|
108 |
generation = generation[0][input_len:]
|
109 |
|
110 |
description = processor.decode(generation, skip_special_tokens=True)
|
|
|
114 |
return state
|
115 |
|
116 |
|
117 |
+
def caption_image(state: State) -> State:
|
118 |
# image is PIL
|
119 |
image = state["image"]
|
120 |
|
|
|
147 |
input_len = inputs["input_ids"].shape[-1]
|
148 |
|
149 |
with torch.inference_mode():
|
150 |
+
generation = model.generate(**inputs, max_new_tokens=1000, do_sample=False)
|
151 |
generation = generation[0][input_len:]
|
152 |
|
153 |
caption = processor.decode(generation, skip_special_tokens=True)
|
app.py
CHANGED
@@ -12,6 +12,7 @@ def process_and_display(image, voice):
|
|
12 |
|
13 |
# Run the graph
|
14 |
result = graph.invoke(state)
|
|
|
15 |
|
16 |
# Return the caption and description
|
17 |
return result["caption"], result["description"]
|
|
|
12 |
|
13 |
# Run the graph
|
14 |
result = graph.invoke(state)
|
15 |
+
print(result)
|
16 |
|
17 |
# Return the caption and description
|
18 |
return result["caption"], result["description"]
|