RichardErkhov commited on
Commit
902a52b
1 Parent(s): 098124b

uploaded readme

Browse files
Files changed (1) hide show
  1. README.md +165 -0
README.md ADDED
@@ -0,0 +1,165 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Quantization made by Richard Erkhov.
2
+
3
+ [Github](https://github.com/RichardErkhov)
4
+
5
+ [Discord](https://discord.gg/pvy7H8DZMG)
6
+
7
+ [Request more models](https://github.com/RichardErkhov/quant_request)
8
+
9
+
10
+ Platypus2-70B - GGUF
11
+ - Model creator: https://huggingface.co/garage-bAInd/
12
+ - Original model: https://huggingface.co/garage-bAInd/Platypus2-70B/
13
+
14
+
15
+ | Name | Quant method | Size |
16
+ | ---- | ---- | ---- |
17
+ | [Platypus2-70B.Q2_K.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.Q2_K.gguf) | Q2_K | 23.71GB |
18
+ | [Platypus2-70B.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.IQ3_XS.gguf) | IQ3_XS | 26.37GB |
19
+ | [Platypus2-70B.IQ3_S.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.IQ3_S.gguf) | IQ3_S | 27.86GB |
20
+ | [Platypus2-70B.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.Q3_K_S.gguf) | Q3_K_S | 27.86GB |
21
+ | [Platypus2-70B.IQ3_M.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.IQ3_M.gguf) | IQ3_M | 28.82GB |
22
+ | [Platypus2-70B.Q3_K.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.Q3_K.gguf) | Q3_K | 30.99GB |
23
+ | [Platypus2-70B.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.Q3_K_M.gguf) | Q3_K_M | 30.99GB |
24
+ | [Platypus2-70B.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.Q3_K_L.gguf) | Q3_K_L | 33.67GB |
25
+ | [Platypus2-70B.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.IQ4_XS.gguf) | IQ4_XS | 34.64GB |
26
+ | [Platypus2-70B.Q4_0.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.Q4_0.gguf) | Q4_0 | 36.2GB |
27
+ | [Platypus2-70B.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.IQ4_NL.gguf) | IQ4_NL | 36.55GB |
28
+ | [Platypus2-70B.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/blob/main/Platypus2-70B.Q4_K_S.gguf) | Q4_K_S | 36.55GB |
29
+ | [Platypus2-70B.Q4_K.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/tree/main/) | Q4_K | 38.58GB |
30
+ | [Platypus2-70B.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/tree/main/) | Q4_K_M | 38.58GB |
31
+ | [Platypus2-70B.Q4_1.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/tree/main/) | Q4_1 | 40.2GB |
32
+ | [Platypus2-70B.Q5_0.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/tree/main/) | Q5_0 | 44.2GB |
33
+ | [Platypus2-70B.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/tree/main/) | Q5_K_S | 44.2GB |
34
+ | [Platypus2-70B.Q5_K.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/tree/main/) | Q5_K | 45.41GB |
35
+ | [Platypus2-70B.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/tree/main/) | Q5_K_M | 45.41GB |
36
+ | [Platypus2-70B.Q5_1.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/tree/main/) | Q5_1 | 48.2GB |
37
+ | [Platypus2-70B.Q6_K.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/tree/main/) | Q6_K | 52.7GB |
38
+ | [Platypus2-70B.Q8_0.gguf](https://huggingface.co/RichardErkhov/garage-bAInd_-_Platypus2-70B-gguf/tree/main/) | Q8_0 | 68.26GB |
39
+
40
+
41
+
42
+
43
+ Original model description:
44
+ ---
45
+ license: cc-by-nc-sa-4.0
46
+ language:
47
+ - en
48
+ datasets:
49
+ - garage-bAInd/Open-Platypus
50
+ ---
51
+
52
+ # Platypus2-70B
53
+
54
+ Platypus-70B is an instruction fine-tuned model based on the LLaMa2-70B transformer architecture.
55
+
56
+ ![Platty](./Best_Platty_small.jpeg)
57
+
58
+ ### Model Details
59
+
60
+ * **Trained by**: Cole Hunter & Ariel Lee
61
+ * **Model type:** **Platypus2-70B** is an auto-regressive language model based on the LLaMA2 transformer architecture.
62
+ * **Language(s)**: English
63
+ * **License for base weights**: Non-Commercial Creative Commons license ([CC BY-NC-4.0](https://creativecommons.org/licenses/by-nc/4.0/))
64
+
65
+ ### Prompt Template
66
+ ```
67
+ ### Instruction:
68
+
69
+ <prompt> (without the <>)
70
+
71
+ ### Response:
72
+ ```
73
+
74
+ ### Training Dataset
75
+
76
+ `garage-bAInd/Platypus2-70B` trained using STEM and logic based dataset [`garage-bAInd/Open-Platypus`](https://huggingface.co/datasets/garage-bAInd/Open-Platypus).
77
+
78
+ Please see our [paper](https://arxiv.org/abs/2308.07317) and [project webpage](https://platypus-llm.github.io) for additional information.
79
+
80
+ ### Training Procedure
81
+
82
+ `garage-bAInd/Platypus2-70B` was instruction fine-tuned using LoRA on 8 A100 80GB. For training details and inference instructions please see the [Platypus](https://github.com/arielnlee/Platypus) GitHub repo.
83
+
84
+ ### Reproducing Evaluation Results
85
+
86
+ Install LM Evaluation Harness:
87
+ ```
88
+ # clone repository
89
+ git clone https://github.com/EleutherAI/lm-evaluation-harness.git
90
+ # check out the correct commit
91
+ git checkout b281b0921b636bc36ad05c0b0b0763bd6dd43463
92
+ # change to repo directory
93
+ cd lm-evaluation-harness
94
+ # install
95
+ pip install -e .
96
+ ```
97
+ Each task was evaluated on a single A100 80GB GPU.
98
+
99
+ ARC:
100
+ ```
101
+ python main.py --model hf-causal-experimental --model_args pretrained=garage-bAInd/Platypus2-70B --tasks arc_challenge --batch_size 1 --no_cache --write_out --output_path results/Platypus2-70B/arc_challenge_25shot.json --device cuda --num_fewshot 25
102
+ ```
103
+
104
+ HellaSwag:
105
+ ```
106
+ python main.py --model hf-causal-experimental --model_args pretrained=garage-bAInd/Platypus2-70B --tasks hellaswag --batch_size 1 --no_cache --write_out --output_path results/Platypus2-70B/hellaswag_10shot.json --device cuda --num_fewshot 10
107
+ ```
108
+
109
+ MMLU:
110
+ ```
111
+ python main.py --model hf-causal-experimental --model_args pretrained=garage-bAInd/Platypus2-70B --tasks hendrycksTest-* --batch_size 1 --no_cache --write_out --output_path results/Platypus2-70B/mmlu_5shot.json --device cuda --num_fewshot 5
112
+ ```
113
+
114
+ TruthfulQA:
115
+ ```
116
+ python main.py --model hf-causal-experimental --model_args pretrained=garage-bAInd/Platypus2-70B --tasks truthfulqa_mc --batch_size 1 --no_cache --write_out --output_path results/Platypus2-70B/truthfulqa_0shot.json --device cuda
117
+ ```
118
+ ### Limitations and bias
119
+
120
+ Llama 2 and fine-tuned variants are a new technology that carries risks with use. Testing conducted to date has been in English, and has not covered, nor could it cover all scenarios. For these reasons, as with all LLMs, Llama 2 and any fine-tuned varient's potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 2 variants, developers should perform safety testing and tuning tailored to their specific applications of the model.
121
+
122
+ Please see the Responsible Use Guide available at https://ai.meta.com/llama/responsible-use-guide/
123
+
124
+ ### Citations
125
+ ```bibtex
126
+ @article{platypus2023,
127
+ title={Platypus: Quick, Cheap, and Powerful Refinement of LLMs},
128
+ author={Ariel N. Lee and Cole J. Hunter and Nataniel Ruiz},
129
+ booktitle={arXiv preprint arxiv:2308.07317},
130
+ year={2023}
131
+ }
132
+ ```
133
+ ```bibtex
134
+ @misc{touvron2023llama,
135
+ title={Llama 2: Open Foundation and Fine-Tuned Chat Models},
136
+ author={Hugo Touvron and Louis Martin and Kevin Stone and Peter Albert and Amjad Almahairi and Yasmine Babaei and Nikolay Bashlykov year={2023},
137
+ eprint={2307.09288},
138
+ archivePrefix={arXiv},
139
+ }
140
+ ```
141
+ ```bibtex
142
+ @inproceedings{
143
+ hu2022lora,
144
+ title={Lo{RA}: Low-Rank Adaptation of Large Language Models},
145
+ author={Edward J Hu and Yelong Shen and Phillip Wallis and Zeyuan Allen-Zhu and Yuanzhi Li and Shean Wang and Lu Wang and Weizhu Chen},
146
+ booktitle={International Conference on Learning Representations},
147
+ year={2022},
148
+ url={https://openreview.net/forum?id=nZeVKeeFYf9}
149
+ }
150
+ ```
151
+ # [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
152
+ Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_garage-bAInd__Platypus2-70B)
153
+
154
+ | Metric | Value |
155
+ |-----------------------|---------------------------|
156
+ | Avg. | 64.16 |
157
+ | ARC (25-shot) | 70.65 |
158
+ | HellaSwag (10-shot) | 87.15 |
159
+ | MMLU (5-shot) | 70.08 |
160
+ | TruthfulQA (0-shot) | 52.37 |
161
+ | Winogrande (5-shot) | 84.37 |
162
+ | GSM8K (5-shot) | 33.06 |
163
+ | DROP (3-shot) | 51.41 |
164
+
165
+