pankajmathur commited on
Commit
0ad4395
1 Parent(s): 1889462

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +120 -1
README.md CHANGED
@@ -3,5 +3,124 @@ language:
3
  - en
4
  library_name: transformers
5
  license: llama2
 
 
 
 
 
 
6
  ---
7
- LlaMA-2 License, more details coming soon...
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  - en
4
  library_name: transformers
5
  license: llama2
6
+ datasets:
7
+ - pankajmathur/orca_mini_v1_dataset
8
+ - pankajmathur/dolly-v2_orca
9
+ - pankajmathur/WizardLM_Orca
10
+ - pankajmathur/alpaca_orca
11
+ - ehartford/dolphin
12
  ---
13
+ # model_009
14
+
15
+ A Llama2-70b model trained on Orca Style datasets.
16
+
17
+ **P.S. If you're interested to collaborate, please connect with me at www.linkedin.com/in/pankajam**
18
+
19
+ ## Evaluation
20
+
21
+ We evaluated model_009 on a wide range of tasks using [Language Model Evaluation Harness](https://github.com/EleutherAI/lm-evaluation-harness) from EleutherAI.
22
+
23
+ Here are the results on metrics used by [HuggingFaceH4 Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
24
+
25
+ |||||
26
+ |:------:|:--------:|:-------:|:--------:|
27
+ |**Task**|**Metric**|**Value**|**Stderr**|
28
+ |*arc_challenge*|acc_norm|0.6843|0.0141|
29
+ |*hellaswag*|acc_norm|0.8671|0.0038|
30
+ |*mmlu*|acc_norm|0.6931|0.0351|
31
+ |*truthfulqa_mc*|mc2|0.5718|0.0157|
32
+ |**Total Average**|-|**0.7041**||
33
+
34
+
35
+ ## Example Usage
36
+
37
+ Here is the prompt format
38
+
39
+ ```
40
+ ### System:
41
+ You are an AI assistant that follows instruction extremely well. Help as much as you can.
42
+
43
+ ### User:
44
+ Tell me about Orcas.
45
+
46
+ ### Assistant:
47
+
48
+ ```
49
+
50
+ Below shows a code example on how to use this model
51
+
52
+ ```python
53
+ import torch
54
+ from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
55
+
56
+ tokenizer = AutoTokenizer.from_pretrained("pankajmathur/model_009")
57
+ model = AutoModelForCausalLM.from_pretrained(
58
+ "pankajmathur/model_009",
59
+ torch_dtype=torch.float16,
60
+ load_in_8bit=True,
61
+ low_cpu_mem_usage=True,
62
+ device_map="auto"
63
+ )
64
+ system_prompt = "### System:\nYou are an AI assistant that follows instruction extremely well. Help as much as you can.\n\n"
65
+
66
+ #generate text steps
67
+ instruction = "Tell me about Orcas."
68
+ prompt = f"{system_prompt}### User: {instruction}\n\n### Assistant:\n"
69
+ inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
70
+ output = model.generate(**inputs, do_sample=True, top_p=0.95, top_k=0, max_new_tokens=4096)
71
+
72
+ print(tokenizer.decode(output[0], skip_special_tokens=True))
73
+
74
+ ```
75
+
76
+
77
+ #### Limitations & Biases:
78
+
79
+ While this model aims for accuracy, it can occasionally produce inaccurate or misleading results.
80
+
81
+ Despite diligent efforts in refining the pretraining data, there remains a possibility for the generation of inappropriate, biased, or offensive content.
82
+
83
+ Exercise caution and cross-check information when necessary.
84
+
85
+
86
+
87
+ ### Citiation:
88
+
89
+ Please kindly cite using the following BibTeX:
90
+
91
+ ```
92
+ @misc{model_009,
93
+ author = {Pankaj Mathur},
94
+ title = {model_009: An Orca Style Llama2-70b model},
95
+ month = {august},
96
+ year = {2023},
97
+ publisher = {HuggingFace},
98
+ journal = {HuggingFace repository},
99
+ howpublished = {\url{https://https://huggingface.co/pankajmathur/model_009},
100
+ }
101
+ ```
102
+
103
+ ```
104
+ @misc{mukherjee2023orca,
105
+ title={Orca: Progressive Learning from Complex Explanation Traces of GPT-4},
106
+ author={Subhabrata Mukherjee and Arindam Mitra and Ganesh Jawahar and Sahaj Agarwal and Hamid Palangi and Ahmed Awadallah},
107
+ year={2023},
108
+ eprint={2306.02707},
109
+ archivePrefix={arXiv},
110
+ primaryClass={cs.CL}
111
+ }
112
+ ```
113
+
114
+ ```
115
+ @software{touvron2023llama2,
116
+ title={Llama 2: Open Foundation and Fine-Tuned Chat Models},
117
+ author={Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava,
118
+ Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu, Wenyin Fu, Brian Fuller,
119
+ Cynthia Gao, Vedanuj Goswami, Naman Goyal, Anthony Hartshorn, Saghar Hosseini, Rui Hou, Hakan Inan, Marcin Kardas, Viktor Kerkez Madian Khabsa, Isabel Kloumann,
120
+ Artem Korenev, Punit Singh Koura, Marie-Anne Lachaux, Thibaut Lavril, Jenya Lee, Diana Liskovich, Yinghai Lu, Yuning Mao, Xavier Martinet, Todor Mihaylov,
121
+ Pushkar Mishra, Igor Molybog, Yixin Nie, Andrew Poulton, Jeremy Reizenstein, Rashi Rungta, Kalyan Saladi, Alan Schelten, Ruan Silva, Eric Michael Smith,
122
+ Ranjan Subramanian, Xiaoqing Ellen Tan, Binh Tang, Ross Taylor, Adina Williams, Jian Xiang Kuan, Puxin Xu , Zheng Yan, Iliyan Zarov, Yuchen Zhang, Angela Fan,
123
+ Melanie Kambadur, Sharan Narang, Aurelien Rodriguez, Robert Stojnic, Sergey Edunov, Thomas Scialom},
124
+ year={2023}
125
+ }
126
+ ```