Hugging Face's logo

RichardErkhov
/
UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf

GGUF
conversational
Model card Files Community
RichardErkhov commited on
Commit
fa0dced
·
verified ·
1 Parent(s): 8ee552c

uploaded readme

Browse files
Files changed (1) hide show
  1. README.md +119 -0
README.md ADDED
@@ -0,0 +1,119 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ Gemma-2-9B-It-SPPO-Iter2 - GGUF
11
+ - Model creator: https://huggingface.co/UCLA-AGI/
12
+ - Original model: https://huggingface.co/UCLA-AGI/Gemma-2-9B-It-SPPO-Iter2/
13
+
14
+
15
+ | Name | Quant method | Size |
16
+ | ---- | ---- | ---- |
17
+ | [Gemma-2-9B-It-SPPO-Iter2.Q2_K.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q2_K.gguf) | Q2_K | 3.54GB |
18
+ | [Gemma-2-9B-It-SPPO-Iter2.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.IQ3_XS.gguf) | IQ3_XS | 3.86GB |
19
+ | [Gemma-2-9B-It-SPPO-Iter2.IQ3_S.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.IQ3_S.gguf) | IQ3_S | 4.04GB |
20
+ | [Gemma-2-9B-It-SPPO-Iter2.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q3_K_S.gguf) | Q3_K_S | 4.04GB |
21
+ | [Gemma-2-9B-It-SPPO-Iter2.IQ3_M.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.IQ3_M.gguf) | IQ3_M | 4.19GB |
22
+ | [Gemma-2-9B-It-SPPO-Iter2.Q3_K.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q3_K.gguf) | Q3_K | 4.43GB |
23
+ | [Gemma-2-9B-It-SPPO-Iter2.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q3_K_M.gguf) | Q3_K_M | 4.43GB |
24
+ | [Gemma-2-9B-It-SPPO-Iter2.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q3_K_L.gguf) | Q3_K_L | 4.78GB |
25
+ | [Gemma-2-9B-It-SPPO-Iter2.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.IQ4_XS.gguf) | IQ4_XS | 4.86GB |
26
+ | [Gemma-2-9B-It-SPPO-Iter2.Q4_0.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q4_0.gguf) | Q4_0 | 5.07GB |
27
+ | [Gemma-2-9B-It-SPPO-Iter2.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.IQ4_NL.gguf) | IQ4_NL | 5.1GB |
28
+ | [Gemma-2-9B-It-SPPO-Iter2.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q4_K_S.gguf) | Q4_K_S | 5.1GB |
29
+ | [Gemma-2-9B-It-SPPO-Iter2.Q4_K.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q4_K.gguf) | Q4_K | 5.37GB |
30
+ | [Gemma-2-9B-It-SPPO-Iter2.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q4_K_M.gguf) | Q4_K_M | 5.37GB |
31
+ | [Gemma-2-9B-It-SPPO-Iter2.Q4_1.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q4_1.gguf) | Q4_1 | 5.55GB |
32
+ | [Gemma-2-9B-It-SPPO-Iter2.Q5_0.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q5_0.gguf) | Q5_0 | 6.04GB |
33
+ | [Gemma-2-9B-It-SPPO-Iter2.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q5_K_S.gguf) | Q5_K_S | 6.04GB |
34
+ | [Gemma-2-9B-It-SPPO-Iter2.Q5_K.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q5_K.gguf) | Q5_K | 6.19GB |
35
+ | [Gemma-2-9B-It-SPPO-Iter2.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q5_K_M.gguf) | Q5_K_M | 6.19GB |
36
+ | [Gemma-2-9B-It-SPPO-Iter2.Q5_1.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q5_1.gguf) | Q5_1 | 6.52GB |
37
+ | [Gemma-2-9B-It-SPPO-Iter2.Q6_K.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q6_K.gguf) | Q6_K | 7.07GB |
38
+ | [Gemma-2-9B-It-SPPO-Iter2.Q8_0.gguf](https://huggingface.co/RichardErkhov/UCLA-AGI_-_Gemma-2-9B-It-SPPO-Iter2-gguf/blob/main/Gemma-2-9B-It-SPPO-Iter2.Q8_0.gguf) | Q8_0 | 9.15GB |
39
+
40
+
41
+
42
+
43
+ Original model description:
44
+ ---
45
+ license: gemma
46
+ datasets:
47
+ - openbmb/UltraFeedback
48
+ language:
49
+ - en
50
+ pipeline_tag: text-generation
51
+ ---
52
+ Self-Play Preference Optimization for Language Model Alignment (https://arxiv.org/abs/2405.00675)
53
+
54
+ # Gemma-2-9B-It-SPPO-Iter2
55
+
56
+ This model was developed using [Self-Play Preference Optimization](https://arxiv.org/abs/2405.00675) at iteration 2, based on the [google/gemma-2-9b-it](https://huggingface.co/google/gemma-2-9b-it) architecture as starting point. We utilized the prompt sets from the [openbmb/UltraFeedback](https://huggingface.co/datasets/openbmb/UltraFeedback) dataset, splited to 3 parts for 3 iterations by [snorkelai/Snorkel-Mistral-PairRM-DPO-Dataset](https://huggingface.co/datasets/snorkelai/Snorkel-Mistral-PairRM-DPO-Dataset). All responses used are synthetic.
57
+
58
+ **Terms of Use**: [Terms](https://www.kaggle.com/models/google/gemma/license/consent/verify/huggingface?returnModelRepoId=google/gemma-2-9b-it)
59
+
60
+
61
+ ## Links to Other Models
62
+ - [Gemma-2-9B-It-SPPO-Iter1](https://huggingface.co/UCLA-AGI/Gemma-2-9B-It-SPPO-Iter1)
63
+ - [Gemma-2-9B-It-SPPO-Iter2](https://huggingface.co/UCLA-AGI/Gemma-2-9B-It-SPPO-Iter2)
64
+ - [Gemma-2-9B-It-SPPO-Iter3](https://huggingface.co/UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3)
65
+
66
+ ### Model Description
67
+
68
+ - Model type: A 8B parameter GPT-like model fine-tuned on synthetic datasets.
69
+ - Language(s) (NLP): Primarily English
70
+ - License: Apache-2.0
71
+ - Finetuned from model: google/gemma-2-9b-it
72
+
73
+
74
+ ## [AlpacaEval Leaderboard Evaluation Results](https://tatsu-lab.github.io/alpaca_eval/)
75
+
76
+
77
+ | Model | LC. Win Rate | Win Rate | Avg. Length |
78
+ |-------------------------------------------|:------------:|:--------:|:-----------:|
79
+ |[Llama-3-8B-SPPO Iter1](https://huggingface.co/UCLA-AGI/Gemma-2-9B-It-SPPO-Iter1) |48.70 |40.76 | 1669
80
+ |[Llama-3-8B-SPPO Iter2](https://huggingface.co/UCLA-AGI/Gemma-2-9B-It-SPPO-Iter2) |50.93 | 44.64 | 1759
81
+ |[Llama-3-8B-SPPO Iter3](https://huggingface.co/UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3) |**53.27** |**47.74** | 1803
82
+
83
+
84
+
85
+
86
+
87
+
88
+ ### Training hyperparameters
89
+ The following hyperparameters were used during training:
90
+
91
+ - learning_rate: 5e-07
92
+ - eta: 1000
93
+ - per_device_train_batch_size: 8
94
+ - gradient_accumulation_steps: 1
95
+ - seed: 42
96
+ - distributed_type: deepspeed_zero3
97
+ - num_devices: 8
98
+ - optimizer: RMSProp
99
+ - lr_scheduler_type: linear
100
+ - lr_scheduler_warmup_ratio: 0.1
101
+ - num_train_epochs: 1.0
102
+
103
+
104
+
105
+
106
+ ## Citation
107
+ ```
108
+ @misc{wu2024self,
109
+ title={Self-Play Preference Optimization for Language Model Alignment},
110
+ author={Wu, Yue and Sun, Zhiqing and Yuan, Huizhuo and Ji, Kaixuan and Yang, Yiming and Gu, Quanquan},
111
+ year={2024},
112
+ eprint={2405.00675},
113
+ archivePrefix={arXiv},
114
+ primaryClass={cs.LG}
115
+ }
116
+ ```
117
+
118
+
119
+