Update app.py
Browse files
app.py
CHANGED
|
@@ -2,16 +2,15 @@ import gradio as gr
|
|
| 2 |
import torch
|
| 3 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
| 4 |
|
| 5 |
-
|
| 6 |
tokenizer = AutoTokenizer.from_pretrained(
|
| 7 |
-
'kakaobrain/kogpt', revision='KoGPT6B-ryan1.5b-float16',
|
| 8 |
bos_token='[BOS]', eos_token='[EOS]', unk_token='[UNK]', pad_token='[PAD]', mask_token='[MASK]'
|
| 9 |
)
|
| 10 |
|
| 11 |
model = AutoModelForCausalLM.from_pretrained(
|
| 12 |
-
'kakaobrain/kogpt', revision='KoGPT6B-ryan1.5b-float16',
|
| 13 |
pad_token_id=tokenizer.eos_token_id,
|
| 14 |
-
torch_dtype=
|
| 15 |
).to(device='cpu', non_blocking=True)
|
| 16 |
_ = model.eval()
|
| 17 |
|
|
|
|
| 2 |
import torch
|
| 3 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
| 4 |
|
|
|
|
| 5 |
tokenizer = AutoTokenizer.from_pretrained(
|
| 6 |
+
'kakaobrain/kogpt', revision='KoGPT6B-ryan1.5b-float16', # or float32 version: revision=KoGPT6B-ryan1.5b
|
| 7 |
bos_token='[BOS]', eos_token='[EOS]', unk_token='[UNK]', pad_token='[PAD]', mask_token='[MASK]'
|
| 8 |
)
|
| 9 |
|
| 10 |
model = AutoModelForCausalLM.from_pretrained(
|
| 11 |
+
'kakaobrain/kogpt', revision='KoGPT6B-ryan1.5b-float16', # or float32 version: revision=KoGPT6B-ryan1.5b
|
| 12 |
pad_token_id=tokenizer.eos_token_id,
|
| 13 |
+
torch_dtype='auto', low_cpu_mem_usage=True
|
| 14 |
).to(device='cpu', non_blocking=True)
|
| 15 |
_ = model.eval()
|
| 16 |
|