Spaces:
Sleeping
Sleeping
jwkirchenbauer
commited on
Commit
β’
811d741
1
Parent(s):
7c3b96d
try fp16
Browse files- app.py +4 -2
- demo_watermark.py +20 -4
app.py
CHANGED
@@ -22,8 +22,10 @@ arg_dict = {
|
|
22 |
'demo_public': False,
|
23 |
# 'model_name_or_path': 'facebook/opt-125m',
|
24 |
# 'model_name_or_path': 'facebook/opt-1.3b',
|
25 |
-
'model_name_or_path': 'facebook/opt-2.7b',
|
26 |
-
|
|
|
|
|
27 |
'prompt_max_length': None,
|
28 |
'max_new_tokens': 200,
|
29 |
'generation_seed': 123,
|
|
|
22 |
'demo_public': False,
|
23 |
# 'model_name_or_path': 'facebook/opt-125m',
|
24 |
# 'model_name_or_path': 'facebook/opt-1.3b',
|
25 |
+
# 'model_name_or_path': 'facebook/opt-2.7b',
|
26 |
+
'model_name_or_path': 'facebook/opt-6.7b',
|
27 |
+
'load_fp16' : True,
|
28 |
+
# 'load_fp16' : False,
|
29 |
'prompt_max_length': None,
|
30 |
'max_new_tokens': 200,
|
31 |
'generation_seed': 123,
|
demo_watermark.py
CHANGED
@@ -162,6 +162,12 @@ def parse_args():
|
|
162 |
default=True,
|
163 |
help="Whether to call the torch seed function before both the unwatermarked and watermarked generate calls.",
|
164 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
165 |
args = parser.parse_args()
|
166 |
return args
|
167 |
|
@@ -173,13 +179,19 @@ def load_model(args):
|
|
173 |
if args.is_seq2seq_model:
|
174 |
model = AutoModelForSeq2SeqLM.from_pretrained(args.model_name_or_path)
|
175 |
elif args.is_decoder_only_model:
|
176 |
-
|
|
|
|
|
|
|
177 |
else:
|
178 |
raise ValueError(f"Unknown model type: {args.model_name_or_path}")
|
179 |
|
180 |
if args.use_gpu:
|
181 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
182 |
-
|
|
|
|
|
|
|
183 |
else:
|
184 |
device = "cpu"
|
185 |
model.eval()
|
@@ -314,8 +326,12 @@ def run_gradio(args, model=None, device=None, tokenizer=None):
|
|
314 |
|
315 |
# Top section, greeting and instructions
|
316 |
gr.Markdown("## π§ [A Watermark for Large Language Models](https://arxiv.org/abs/2301.10226) π")
|
317 |
-
gr.
|
318 |
-
|
|
|
|
|
|
|
|
|
319 |
with gr.Accordion("Understanding the output metrics",open=False):
|
320 |
gr.Markdown(
|
321 |
"""
|
|
|
162 |
default=True,
|
163 |
help="Whether to call the torch seed function before both the unwatermarked and watermarked generate calls.",
|
164 |
)
|
165 |
+
parser.add_argument(
|
166 |
+
"--load_fp16",
|
167 |
+
type=str2bool,
|
168 |
+
default=False,
|
169 |
+
help="Whether to run model in float16 precsion.",
|
170 |
+
)
|
171 |
args = parser.parse_args()
|
172 |
return args
|
173 |
|
|
|
179 |
if args.is_seq2seq_model:
|
180 |
model = AutoModelForSeq2SeqLM.from_pretrained(args.model_name_or_path)
|
181 |
elif args.is_decoder_only_model:
|
182 |
+
if args.load_fp16:
|
183 |
+
model = AutoModelForCausalLM.from_pretrained(args.model_name_or_path,torch_dtype=torch.float16, device_map='auto')
|
184 |
+
else:
|
185 |
+
model = AutoModelForCausalLM.from_pretrained(args.model_name_or_path)
|
186 |
else:
|
187 |
raise ValueError(f"Unknown model type: {args.model_name_or_path}")
|
188 |
|
189 |
if args.use_gpu:
|
190 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
191 |
+
if args.load_fp16:
|
192 |
+
pass
|
193 |
+
else:
|
194 |
+
model = model.to(device)
|
195 |
else:
|
196 |
device = "cpu"
|
197 |
model.eval()
|
|
|
326 |
|
327 |
# Top section, greeting and instructions
|
328 |
gr.Markdown("## π§ [A Watermark for Large Language Models](https://arxiv.org/abs/2301.10226) π")
|
329 |
+
with gr.Row():
|
330 |
+
gr.Markdown("![visitor badge](https://visitor-badge.glitch.me/badge?page_id=tomg-group-umd_lm-watermarking)")
|
331 |
+
with gr.Row():
|
332 |
+
gr.Markdown("[jwkirchenbauer/lm-watermarking![](https://badgen.net/badge/icon/GitHub?icon=github&label)](https://github.com/jwkirchenbauer/lm-watermarking)")
|
333 |
+
with gr.Row():
|
334 |
+
gr.Markdown(f"Language model: {args.model_name_or_path}")
|
335 |
with gr.Accordion("Understanding the output metrics",open=False):
|
336 |
gr.Markdown(
|
337 |
"""
|