Pankaj Mathur
commited on
Commit
•
58af173
1
Parent(s):
335b148
Update README.md
Browse files
README.md
CHANGED
@@ -2,11 +2,13 @@
|
|
2 |
language:
|
3 |
- en
|
4 |
library_name: transformers
|
|
|
|
|
5 |
---
|
6 |
|
7 |
# model_42_70b
|
8 |
|
9 |
-
A Llama2-70b model
|
10 |
|
11 |
**I am actively seeking sponsorship and partnership opportunities. If you're interested, please connect with me at www.linkedin.com/in/pankajam.**
|
12 |
|
@@ -31,11 +33,9 @@ Here are the results on metrics used by [HuggingFaceH4 Open LLM Leaderboard](htt
|
|
31 |
Here is the prompt format
|
32 |
|
33 |
```
|
34 |
-
### System:
|
35 |
-
You are an AI assistant that follows instruction extremely well. Help as much as you can.
|
36 |
|
37 |
### User:
|
38 |
-
|
39 |
|
40 |
### Assistant:
|
41 |
|
@@ -55,11 +55,10 @@ model = AutoModelForCausalLM.from_pretrained(
|
|
55 |
low_cpu_mem_usage=True,
|
56 |
device_map="auto"
|
57 |
)
|
58 |
-
system_prompt = "### System:\nYou are an AI assistant that follows instruction extremely well. Help as much as you can.\n\n"
|
59 |
|
60 |
#generate text steps
|
61 |
-
instruction = "
|
62 |
-
prompt = f"
|
63 |
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
|
64 |
output = model.generate(**inputs, do_sample=True, top_p=0.95, top_k=0, max_new_tokens=4096)
|
65 |
|
@@ -85,7 +84,7 @@ Please kindly cite using the following BibTeX:
|
|
85 |
```
|
86 |
@misc{model_42_70b,
|
87 |
author = {Pankaj Mathur},
|
88 |
-
title = {model_42_70b:
|
89 |
year = {2023},
|
90 |
publisher = {HuggingFace},
|
91 |
journal = {HuggingFace repository},
|
@@ -94,11 +93,12 @@ Please kindly cite using the following BibTeX:
|
|
94 |
```
|
95 |
|
96 |
```
|
97 |
-
@misc{
|
98 |
-
title={
|
99 |
-
author={
|
|
|
100 |
year={2023},
|
101 |
-
eprint={
|
102 |
archivePrefix={arXiv},
|
103 |
primaryClass={cs.CL}
|
104 |
}
|
|
|
2 |
language:
|
3 |
- en
|
4 |
library_name: transformers
|
5 |
+
datasets:
|
6 |
+
- GAIR/lima
|
7 |
---
|
8 |
|
9 |
# model_42_70b
|
10 |
|
11 |
+
A Llama2-70b model fine tuned on lima datasets, which has < ~1k conversations.
|
12 |
|
13 |
**I am actively seeking sponsorship and partnership opportunities. If you're interested, please connect with me at www.linkedin.com/in/pankajam.**
|
14 |
|
|
|
33 |
Here is the prompt format
|
34 |
|
35 |
```
|
|
|
|
|
36 |
|
37 |
### User:
|
38 |
+
Write a stand-up skit in the style of George Carlin that ridicules Pacific Gas and Electric.
|
39 |
|
40 |
### Assistant:
|
41 |
|
|
|
55 |
low_cpu_mem_usage=True,
|
56 |
device_map="auto"
|
57 |
)
|
|
|
58 |
|
59 |
#generate text steps
|
60 |
+
instruction = "Write a stand-up skit in the style of George Carlin that ridicules Pacific Gas and Electric."
|
61 |
+
prompt = f"### User: {instruction}\n\n### Assistant:\n"
|
62 |
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
|
63 |
output = model.generate(**inputs, do_sample=True, top_p=0.95, top_k=0, max_new_tokens=4096)
|
64 |
|
|
|
84 |
```
|
85 |
@misc{model_42_70b,
|
86 |
author = {Pankaj Mathur},
|
87 |
+
title = {model_42_70b: A LIMA style Llama2-70b model},
|
88 |
year = {2023},
|
89 |
publisher = {HuggingFace},
|
90 |
journal = {HuggingFace repository},
|
|
|
93 |
```
|
94 |
|
95 |
```
|
96 |
+
@misc{ChuntingZhou,
|
97 |
+
title={LIMA: Less Is More for Alignment},
|
98 |
+
author={Chunting Zhou, Pengfei Liu, Puxin Xu, Srini Iyer, Jiao Sun, Yuning Mao, Xuezhe Ma, Avia Efrat, Ping Yu, Lili Yu,
|
99 |
+
Susan Zhang, Gargi Ghosh, Mike Lewis, Luke Zettlemoyer, Omer Levy},
|
100 |
year={2023},
|
101 |
+
eprint={2305.11206},
|
102 |
archivePrefix={arXiv},
|
103 |
primaryClass={cs.CL}
|
104 |
}
|