Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
datasets:
|
4 |
+
- OpenAssistant/oasst1
|
5 |
+
- erfanzar/CC-H2OAI-OASST-1-TRAIN
|
6 |
+
- erfanzar/CC-OASST-1-TRAIN
|
7 |
+
language:
|
8 |
+
- en
|
9 |
+
- fr
|
10 |
+
- fa
|
11 |
+
- nl
|
12 |
+
metrics:
|
13 |
+
- bertscore
|
14 |
+
pipeline_tag: text-generation
|
15 |
+
---
|
16 |
+
|
17 |
+
|
18 |
+
## Hello community
|
19 |
+
|
20 |
+
this model is only 1B but you can call it somehow an SOTA
|
21 |
+
|
22 |
+
|
23 |
+
this model can also run on 4 GB GPU RAM and know dialogs as well
|
24 |
+
|
25 |
+
## Usage Code
|
26 |
+
|
27 |
+
```python
|
28 |
+
|
29 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
30 |
+
from IPython.display import clear_output
|
31 |
+
import textwrap
|
32 |
+
|
33 |
+
|
34 |
+
tokenizer = AutoTokenizer.from_pretrained("erfanzar/PGT-1B-2EP")
|
35 |
+
|
36 |
+
model = AutoModelForCausalLM.from_pretrained("erfanzar/PGT-1B-2EP",device_map='auto',load_in_8bit=True)
|
37 |
+
|
38 |
+
|
39 |
+
verify_text = lambda txt : '\n'.join([textwrap.fill(txt, width=140) for txt in txt.split('\n')])
|
40 |
+
|
41 |
+
|
42 |
+
def ppp(text:str):
|
43 |
+
"""
|
44 |
+
pre processing prompt
|
45 |
+
"""
|
46 |
+
return f"<|prompter|>{text}<|endoftext|><|assistant|>"
|
47 |
+
|
48 |
+
def generate(text,max_new_tokens:int=512,use_ppp:bool=False,b_pair=False):
|
49 |
+
text = ppp(text) if use_ppp else text
|
50 |
+
|
51 |
+
for i in range(max_new_tokens):
|
52 |
+
enc = tokenizer(text,return_tensors='pt')
|
53 |
+
text_r = text
|
54 |
+
enc = model.generate(**enc,max_new_tokens=1,pad_token_id=0)
|
55 |
+
text = tokenizer.decode(enc[0])
|
56 |
+
if text.endswith(tokenizer.eos_token):
|
57 |
+
break
|
58 |
+
else:
|
59 |
+
yield text[len(text_r):] if b_pair else text
|
60 |
+
|
61 |
+
for v in generate('where is empire building ?',512,True):
|
62 |
+
clear_output(wait=True)
|
63 |
+
print(verify_text(v),end='')
|
64 |
+
|
65 |
+
```
|
66 |
+
|
67 |
+
# Pythia-1B
|
68 |
+
|
69 |
+
## Model Details
|
70 |
+
|
71 |
+
### Pretrained Model
|
72 |
+
|
73 |
+
- Developed by: [EleutherAI](http://eleuther.ai)
|
74 |
+
- Model type: Transformer-based Language Model
|
75 |
+
- FineTuned Languages: English , Persian , French, And Dutch
|
76 |
+
- Learn more: [Pythia's GitHub repository](https://github.com/EleutherAI/pythia) for training procedures, config files, and details on how to use.
|
77 |
+
- Library: [GPT-NeoX](https://github.com/EleutherAI/gpt-neox)
|
78 |
+
- License: [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)
|
79 |
+
|
80 |
+
|
81 |
+
|
82 |
+
## NOTE
|
83 |
+
|
84 |
+
The Pythia Suite is **NOT** intended for deployment. It is not in itself
|
85 |
+
a product and cannot be used for human-facing interactions. For example,
|
86 |
+
the model may generate harmful or offensive text...
|
87 |
+
|
88 |
+
|
89 |
+
and also remember that this model is not good enough for Persian, French, and Dutch at least for this version
|