Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -1,3 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import os
|
2 |
import time
|
3 |
import spaces
|
@@ -36,7 +42,8 @@ h3 {
|
|
36 |
|
37 |
model = AutoModelForCausalLM.from_pretrained(
|
38 |
MODEL_ID,
|
39 |
-
torch_dtype=torch.float16,
|
|
|
40 |
trust_remote_code=True).cuda()
|
41 |
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
|
42 |
|
|
|
1 |
+
import subprocess
|
2 |
+
subprocess.run(
|
3 |
+
'pip install flash-attn --no-build-isolation',
|
4 |
+
env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"},
|
5 |
+
shell=True
|
6 |
+
)
|
7 |
import os
|
8 |
import time
|
9 |
import spaces
|
|
|
42 |
|
43 |
model = AutoModelForCausalLM.from_pretrained(
|
44 |
MODEL_ID,
|
45 |
+
torch_dtype=torch.float16,
|
46 |
+
attn_implementation="flash_attention_2",
|
47 |
trust_remote_code=True).cuda()
|
48 |
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
|
49 |
|