v1
Browse files- .gitignore +3 -0
- app.py +3 -5
- trol/load_trol.py +1 -1
.gitignore
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
.vscode
|
2 |
+
__pycache__
|
3 |
+
*/__pycache__
|
app.py
CHANGED
@@ -38,8 +38,6 @@ model_3_8, tokenizer_3_8 = load_trol(link='TroL-3.8B')
|
|
38 |
# loading model
|
39 |
model_7, tokenizer_7 = load_trol(link='TroL-7B')
|
40 |
|
41 |
-
print()
|
42 |
-
|
43 |
def threading_function(inputs, image_token_number, streamer, device, model, tokenizer, temperature, new_max_token, top_p):
|
44 |
|
45 |
# propagation
|
@@ -85,7 +83,7 @@ def bot_streaming(message, history, link, temperature, new_max_token, top_p):
|
|
85 |
if "3.8B" not in link:
|
86 |
image_token_number = 1225
|
87 |
image = F.interpolate(image.unsqueeze(0), size=(490, 490), mode='bicubic').squeeze(0)
|
88 |
-
inputs = [{'image': image, 'question': message['text']}]
|
89 |
elif len(message['files']) > 1:
|
90 |
raise Exception("No way!")
|
91 |
else:
|
@@ -116,7 +114,7 @@ def bot_streaming(message, history, link, temperature, new_max_token, top_p):
|
|
116 |
|
117 |
# Text decoding
|
118 |
response = output_filtering(generated_text, model)
|
119 |
-
|
120 |
except:
|
121 |
response = "There may be unsupported format: ex) pdf, video, sound. Only supported is a single image in this version."
|
122 |
|
@@ -138,7 +136,7 @@ def bot_streaming(message, history, link, temperature, new_max_token, top_p):
|
|
138 |
yield buffer
|
139 |
|
140 |
demo = gr.ChatInterface(fn=bot_streaming,
|
141 |
-
additional_inputs = [gr.Radio(["3.8B"], label="Size", info="Select one model size", value="3.8B"), gr.Slider(0, 1, 0.9, label="temperature"), gr.Slider(1, 1024, 128, label="new_max_token"), gr.Slider(0, 1, 0.95, label="top_p")],
|
142 |
additional_inputs_accordion="Generation Hyperparameters",
|
143 |
theme=gr.themes.Soft(),
|
144 |
title="TroL",
|
|
|
38 |
# loading model
|
39 |
model_7, tokenizer_7 = load_trol(link='TroL-7B')
|
40 |
|
|
|
|
|
41 |
def threading_function(inputs, image_token_number, streamer, device, model, tokenizer, temperature, new_max_token, top_p):
|
42 |
|
43 |
# propagation
|
|
|
83 |
if "3.8B" not in link:
|
84 |
image_token_number = 1225
|
85 |
image = F.interpolate(image.unsqueeze(0), size=(490, 490), mode='bicubic').squeeze(0)
|
86 |
+
inputs = [{'image': image.to(accel.device), 'question': message['text']}]
|
87 |
elif len(message['files']) > 1:
|
88 |
raise Exception("No way!")
|
89 |
else:
|
|
|
114 |
|
115 |
# Text decoding
|
116 |
response = output_filtering(generated_text, model)
|
117 |
+
|
118 |
except:
|
119 |
response = "There may be unsupported format: ex) pdf, video, sound. Only supported is a single image in this version."
|
120 |
|
|
|
136 |
yield buffer
|
137 |
|
138 |
demo = gr.ChatInterface(fn=bot_streaming,
|
139 |
+
additional_inputs = [gr.Radio(["1.8B", "3.8B"], label="Size", info="Select one model size", value="3.8B"), gr.Slider(0, 1, 0.9, label="temperature"), gr.Slider(1, 1024, 128, label="new_max_token"), gr.Slider(0, 1, 0.95, label="top_p")],
|
140 |
additional_inputs_accordion="Generation Hyperparameters",
|
141 |
theme=gr.themes.Soft(),
|
142 |
title="TroL",
|
trol/load_trol.py
CHANGED
@@ -23,7 +23,7 @@ def load_trol(link):
|
|
23 |
from transformers import LlamaTokenizerFast as TroLTokenizer
|
24 |
bits = 8
|
25 |
path = TROL_3_8B
|
26 |
-
bit_quant_skip = ["vision_model", "
|
27 |
|
28 |
elif link == 'TroL-7B':
|
29 |
from .arch_internlm2.modeling_trol import TroLForCausalLM
|
|
|
23 |
from transformers import LlamaTokenizerFast as TroLTokenizer
|
24 |
bits = 8
|
25 |
path = TROL_3_8B
|
26 |
+
bit_quant_skip = ["vision_model", "vision_proj", "lm_head"]
|
27 |
|
28 |
elif link == 'TroL-7B':
|
29 |
from .arch_internlm2.modeling_trol import TroLForCausalLM
|