Spaces:
Runtime error
Runtime error
jaymojnidar
commited on
Commit
•
e4aabb2
1
Parent(s):
903e4f3
adding my data config peft
Browse files- model.py +17 -9
- requirements.txt +3 -0
model.py
CHANGED
@@ -3,6 +3,7 @@ from threading import Thread
|
|
3 |
from typing import Iterator
|
4 |
|
5 |
import torch
|
|
|
6 |
from transformers import AutoConfig, AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
7 |
from huggingface_hub import login
|
8 |
|
@@ -11,6 +12,8 @@ from huggingface_hub import login
|
|
11 |
#model_id = 'Trelis/Llama-2-7b-chat-hf-sharded-bf16'
|
12 |
model_id = 'jaymojnidar/llama2-finetuned-mydata'
|
13 |
config_model_id = 'jaymojnidar/llama2-finetuned-mydata/adapter_config.json'
|
|
|
|
|
14 |
|
15 |
if torch.cuda.is_available():
|
16 |
tok = os.environ['HF_TOKEN']
|
@@ -21,15 +24,20 @@ if torch.cuda.is_available():
|
|
21 |
#, token="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"
|
22 |
)
|
23 |
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
|
|
|
|
|
|
|
|
|
|
33 |
)
|
34 |
print("Loaded the model!")
|
35 |
else:
|
|
|
3 |
from typing import Iterator
|
4 |
|
5 |
import torch
|
6 |
+
from peft import PeftModel, PeftConfig
|
7 |
from transformers import AutoConfig, AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
8 |
from huggingface_hub import login
|
9 |
|
|
|
12 |
#model_id = 'Trelis/Llama-2-7b-chat-hf-sharded-bf16'
|
13 |
model_id = 'jaymojnidar/llama2-finetuned-mydata'
|
14 |
config_model_id = 'jaymojnidar/llama2-finetuned-mydata/adapter_config.json'
|
15 |
+
model_type = 'PEFT'
|
16 |
+
|
17 |
|
18 |
if torch.cuda.is_available():
|
19 |
tok = os.environ['HF_TOKEN']
|
|
|
24 |
#, token="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"
|
25 |
)
|
26 |
|
27 |
+
if model_type == 'PEFT':
|
28 |
+
config = PeftConfig.from_pretrained("jaymojnidar/llama2-finetuned-mydata")
|
29 |
+
model = AutoModelForCausalLM.from_pretrained("Trelis/Llama-2-7b-chat-hf-sharded-bf16")
|
30 |
+
model = PeftModel.from_pretrained(model, "jaymojnidar/llama2-finetuned-mydata")
|
31 |
+
else:
|
32 |
+
config = AutoConfig.from_pretrained(model_id, use_auth_token=True)
|
33 |
+
config.pretraining_tp = 1
|
34 |
+
model = AutoModelForCausalLM.from_pretrained(
|
35 |
+
model_id,
|
36 |
+
config=config,
|
37 |
+
torch_dtype=torch.float16,
|
38 |
+
#load_in_4bit=True,
|
39 |
+
device_map='auto',
|
40 |
+
use_auth_token=True
|
41 |
)
|
42 |
print("Loaded the model!")
|
43 |
else:
|
requirements.txt
CHANGED
@@ -1,8 +1,11 @@
|
|
1 |
accelerate==0.21.0
|
2 |
bitsandbytes==0.40.2
|
3 |
gradio==3.37.0
|
|
|
4 |
protobuf==3.20.3
|
5 |
scipy==1.11.1
|
6 |
sentencepiece==0.1.99
|
|
|
|
|
7 |
torch==2.0.1
|
8 |
transformers==4.31.0
|
|
|
1 |
accelerate==0.21.0
|
2 |
bitsandbytes==0.40.2
|
3 |
gradio==3.37.0
|
4 |
+
peft==0.5.0
|
5 |
protobuf==3.20.3
|
6 |
scipy==1.11.1
|
7 |
sentencepiece==0.1.99
|
8 |
+
#--extra-index-url https://download.pytorch.org/whl/cu113
|
9 |
+
#torch
|
10 |
torch==2.0.1
|
11 |
transformers==4.31.0
|