Hugging Face's logo

RichardErkhov
/
RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf

GGUF
Inference Endpoints
conversational
Model card Files Community
RichardErkhov commited on
Commit
a828d1b
•
1 Parent(s): f504c67

uploaded readme

Browse files
Files changed (1) hide show
  1. README.md +114 -0
README.md ADDED
@@ -0,0 +1,114 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5 - GGUF
11
+ - Model creator: https://huggingface.co/RyanYr/
12
+ - Original model: https://huggingface.co/RyanYr/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5/
13
+
14
+
15
+ | Name | Quant method | Size |
16
+ | ---- | ---- | ---- |
17
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q2_K.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q2_K.gguf) | Q2_K | 1.39GB |
18
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.IQ3_XS.gguf) | IQ3_XS | 1.53GB |
19
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.IQ3_S.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.IQ3_S.gguf) | IQ3_S | 1.59GB |
20
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q3_K_S.gguf) | Q3_K_S | 1.59GB |
21
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.IQ3_M.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.IQ3_M.gguf) | IQ3_M | 1.65GB |
22
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q3_K.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q3_K.gguf) | Q3_K | 1.73GB |
23
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q3_K_M.gguf) | Q3_K_M | 1.73GB |
24
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q3_K_L.gguf) | Q3_K_L | 1.85GB |
25
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.IQ4_XS.gguf) | IQ4_XS | 1.91GB |
26
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q4_0.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q4_0.gguf) | Q4_0 | 1.99GB |
27
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.IQ4_NL.gguf) | IQ4_NL | 2.0GB |
28
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q4_K_S.gguf) | Q4_K_S | 2.0GB |
29
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q4_K.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q4_K.gguf) | Q4_K | 2.09GB |
30
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q4_K_M.gguf) | Q4_K_M | 2.09GB |
31
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q4_1.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q4_1.gguf) | Q4_1 | 2.18GB |
32
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q5_0.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q5_0.gguf) | Q5_0 | 2.37GB |
33
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q5_K_S.gguf) | Q5_K_S | 2.37GB |
34
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q5_K.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q5_K.gguf) | Q5_K | 2.41GB |
35
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q5_K_M.gguf) | Q5_K_M | 2.41GB |
36
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q5_1.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q5_1.gguf) | Q5_1 | 2.55GB |
37
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q6_K.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q6_K.gguf) | Q6_K | 2.76GB |
38
+ | [self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q8_0.gguf](https://huggingface.co/RichardErkhov/RyanYr_-_self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5-gguf/blob/main/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5.Q8_0.gguf) | Q8_0 | 3.58GB |
39
+
40
+
41
+
42
+
43
+ Original model description:
44
+ ---
45
+ base_model: RyanYr/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter4
46
+ library_name: transformers
47
+ model_name: self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5
48
+ tags:
49
+ - generated_from_trainer
50
+ - trl
51
+ - dpo
52
+ licence: license
53
+ ---
54
+
55
+ # Model Card for self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5
56
+
57
+ This model is a fine-tuned version of [RyanYr/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter4](https://huggingface.co/RyanYr/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter4).
58
+ It has been trained using [TRL](https://github.com/huggingface/trl).
59
+
60
+ ## Quick start
61
+
62
+ ```python
63
+ from transformers import pipeline
64
+
65
+ question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
66
+ generator = pipeline("text-generation", model="RyanYr/self-correct_Llama-3.2-3B-Instruct_metaMathQA_dpo_iter5", device="cuda")
67
+ output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
68
+ print(output["generated_text"])
69
+ ```
70
+
71
+ ## Training procedure
72
+
73
+ [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/yyr/huggingface/runs/829uv72l)
74
+
75
+ This model was trained with DPO, a method introduced in [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://huggingface.co/papers/2305.18290).
76
+
77
+ ### Framework versions
78
+
79
+ - TRL: 0.12.0.dev0
80
+ - Transformers: 4.45.2
81
+ - Pytorch: 2.4.0
82
+ - Datasets: 3.0.1
83
+ - Tokenizers: 0.20.1
84
+
85
+ ## Citations
86
+
87
+ Cite DPO as:
88
+
89
+ ```bibtex
90
+ @inproceedings{rafailov2023direct,
91
+ title = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
92
+ author = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
93
+ year = 2023,
94
+ booktitle = {Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023},
95
+ url = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
96
+ editor = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
97
+ }
98
+ ```
99
+
100
+ Cite TRL as:
101
+
102
+ ```bibtex
103
+ @misc{vonwerra2022trl,
104
+ title = {{TRL: Transformer Reinforcement Learning}},
105
+ author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
106
+ year = 2020,
107
+ journal = {GitHub repository},
108
+ publisher = {GitHub},
109
+ howpublished = {\url{https://github.com/huggingface/trl}}
110
+ }
111
+ ```
112
+
113
+
114
+ Additional thanks to @nicoboss for giving me access to his private supercomputer, enabling me to provide many more quants, at much higher speed, than I would otherwise be able to.