Hugging Face's logo

RichardErkhov
/
MiniLLM_-_MiniPLM-llama3.1-212M-gguf

GGUF
Inference Endpoints
Model card Files Community
RichardErkhov commited on
Commit
307e2f7
·
verified ·
1 Parent(s): a188581

uploaded readme

Browse files
Files changed (1) hide show
  1. README.md +91 -0
README.md ADDED
@@ -0,0 +1,91 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ MiniPLM-llama3.1-212M - GGUF
11
+ - Model creator: https://huggingface.co/MiniLLM/
12
+ - Original model: https://huggingface.co/MiniLLM/MiniPLM-llama3.1-212M/
13
+
14
+
15
+ | Name | Quant method | Size |
16
+ | ---- | ---- | ---- |
17
+ | [MiniPLM-llama3.1-212M.Q2_K.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q2_K.gguf) | Q2_K | 0.12GB |
18
+ | [MiniPLM-llama3.1-212M.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.IQ3_XS.gguf) | IQ3_XS | 0.13GB |
19
+ | [MiniPLM-llama3.1-212M.IQ3_S.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.IQ3_S.gguf) | IQ3_S | 0.13GB |
20
+ | [MiniPLM-llama3.1-212M.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q3_K_S.gguf) | Q3_K_S | 0.13GB |
21
+ | [MiniPLM-llama3.1-212M.IQ3_M.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.IQ3_M.gguf) | IQ3_M | 0.13GB |
22
+ | [MiniPLM-llama3.1-212M.Q3_K.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q3_K.gguf) | Q3_K | 0.13GB |
23
+ | [MiniPLM-llama3.1-212M.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q3_K_M.gguf) | Q3_K_M | 0.13GB |
24
+ | [MiniPLM-llama3.1-212M.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q3_K_L.gguf) | Q3_K_L | 0.14GB |
25
+ | [MiniPLM-llama3.1-212M.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.IQ4_XS.gguf) | IQ4_XS | 0.14GB |
26
+ | [MiniPLM-llama3.1-212M.Q4_0.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q4_0.gguf) | Q4_0 | 0.14GB |
27
+ | [MiniPLM-llama3.1-212M.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.IQ4_NL.gguf) | IQ4_NL | 0.14GB |
28
+ | [MiniPLM-llama3.1-212M.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q4_K_S.gguf) | Q4_K_S | 0.14GB |
29
+ | [MiniPLM-llama3.1-212M.Q4_K.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q4_K.gguf) | Q4_K | 0.15GB |
30
+ | [MiniPLM-llama3.1-212M.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q4_K_M.gguf) | Q4_K_M | 0.15GB |
31
+ | [MiniPLM-llama3.1-212M.Q4_1.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q4_1.gguf) | Q4_1 | 0.15GB |
32
+ | [MiniPLM-llama3.1-212M.Q5_0.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q5_0.gguf) | Q5_0 | 0.16GB |
33
+ | [MiniPLM-llama3.1-212M.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q5_K_S.gguf) | Q5_K_S | 0.16GB |
34
+ | [MiniPLM-llama3.1-212M.Q5_K.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q5_K.gguf) | Q5_K | 0.16GB |
35
+ | [MiniPLM-llama3.1-212M.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q5_K_M.gguf) | Q5_K_M | 0.16GB |
36
+ | [MiniPLM-llama3.1-212M.Q5_1.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q5_1.gguf) | Q5_1 | 0.16GB |
37
+ | [MiniPLM-llama3.1-212M.Q6_K.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q6_K.gguf) | Q6_K | 0.17GB |
38
+ | [MiniPLM-llama3.1-212M.Q8_0.gguf](https://huggingface.co/RichardErkhov/MiniLLM_-_MiniPLM-llama3.1-212M-gguf/blob/main/MiniPLM-llama3.1-212M.Q8_0.gguf) | Q8_0 | 0.22GB |
39
+
40
+
41
+
42
+
43
+ Original model description:
44
+ ---
45
+ library_name: transformers
46
+ license: apache-2.0
47
+ datasets:
48
+ - monology/pile-uncopyrighted
49
+ - MiniLLM/pile-diff_samp-qwen_1.8B-qwen_104M-r0.5
50
+ language:
51
+ - en
52
+ metrics:
53
+ - accuracy
54
+ pipeline_tag: text-generation
55
+ ---
56
+
57
+ # MiniPLM-llama3.1-212M
58
+
59
+ [paper](https://arxiv.org/abs/2410.17215) | [code](https://github.com/thu-coai/MiniPLM)
60
+
61
+ **MiniPLM-llama3.1-212M** is a 212M model with the [LLaMA3.1 achitecture](https://arxiv.org/abs/2407.21783) pre-trained from scratch on [the Pile](https://huggingface.co/datasets/monology/pile-uncopyrighted) using the MiniPLM knowledge distillation framework with the [offcial Qwen1.5-1.8B](https://huggingface.co/Qwen/Qwen1.5-1.8B) as the teacher model.
62
+ This model shows the flexibility of the MiniPLM framework in conducting knowledge distillation across model families.
63
+
64
+ We also open-source the [pre-training corpus](https://huggingface.co/datasets/MiniLLM/pile-diff_samp-qwen_1.8B-qwen_104M-r0.5) refined by Difference Sampling in MiniPLM for reproducibility.
65
+
66
+ <p align='left'>
67
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/624ac662102fcdff87be51b9/2BqT0NgkmIXYlktovw9kG.png" width="1000">
68
+ </p>
69
+
70
+ ## Evaluation
71
+
72
+ MiniPLM models achieves better performance given the same computation and scales well across model sizes:
73
+
74
+ <p align='left'>
75
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/624ac662102fcdff87be51b9/EOYzajQcwQFT5PobqL3j0.png" width="1000">
76
+ </p>
77
+
78
+ ## Baseline Models
79
+ + [Conventional Pre-Training](https://huggingface.co/MiniLLM/Pretrain-LLama3.1-130M)
80
+
81
+ ## Citation
82
+
83
+ ```bibtex
84
+ @article{miniplm,
85
+ title={MiniPLM: Knowledge Distillation for Pre-Training Language Models},
86
+ author={Yuxian Gu and Hao Zhou and Fandong Meng and Jie Zhou and Minlie Huang},
87
+ journal={arXiv preprint arXiv:2410.17215},
88
+ year={2024}
89
+ }
90
+ ```
91
+