README.md
6.1 KB · 90 lines · markdown Raw
1 ---
2 base_model: WillHeld/Delphi-25B-SimpleRL-Math
3 language:
4 - en
5 library_name: transformers
6 mradermacher:
7 readme_rev: 1
8 quantized_by: mradermacher
9 tags:
10 - math
11 - reasoning
12 - grpo
13 - reinforcement-learning
14 - simplerl
15 ---
16 ## About
17
18 <!-- ### quantize_version: 2 -->
19 <!-- ### output_tensor_quantised: 1 -->
20 <!-- ### convert_type: hf -->
21 <!-- ### vocab_type: -->
22 <!-- ### tags: nicoboss -->
23 <!-- ### quants: Q2_K IQ3_M Q4_K_S IQ3_XXS Q3_K_M small-IQ4_NL Q4_K_M IQ2_M Q6_K IQ4_XS Q2_K_S IQ1_M Q3_K_S IQ2_XXS Q3_K_L IQ2_XS Q5_K_S IQ2_S IQ1_S Q5_K_M Q4_0 IQ3_XS Q4_1 IQ3_S -->
24 <!-- ### quants_skip: -->
25 <!-- ### skip_mmproj: -->
26 weighted/imatrix quants of https://huggingface.co/WillHeld/Delphi-25B-SimpleRL-Math
27
28 <!-- provided-files -->
29
30 ***For a convenient overview and download list, visit our [model page for this model](https://hf.tst.eu/model#Delphi-25B-SimpleRL-Math-i1-GGUF).***
31
32 static quants are available at https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-GGUF
33 ## Usage
34
35 If you are unsure how to use GGUF files, refer to one of [TheBloke's
36 READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
37 more details, including on how to concatenate multi-part files.
38
39 ## Provided Quants
40
41 (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
42
43 | Link | Type | Size/GB | Notes |
44 |:-----|:-----|--------:|:------|
45 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.imatrix.gguf) | imatrix | 0.1 | imatrix file (for creating your own quants) |
46 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ1_S.gguf) | i1-IQ1_S | 5.8 | for the desperate |
47 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ1_M.gguf) | i1-IQ1_M | 6.3 | mostly desperate |
48 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 7.0 | |
49 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ2_XS.gguf) | i1-IQ2_XS | 7.7 | |
50 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ2_S.gguf) | i1-IQ2_S | 8.2 | |
51 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q2_K_S.gguf) | i1-Q2_K_S | 8.8 | very low quality |
52 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ2_M.gguf) | i1-IQ2_M | 8.8 | |
53 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q2_K.gguf) | i1-Q2_K | 9.5 | IQ3_XXS probably better |
54 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 9.8 | lower quality |
55 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ3_XS.gguf) | i1-IQ3_XS | 10.6 | |
56 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ3_S.gguf) | i1-IQ3_S | 11.1 | beats Q3_K* |
57 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q3_K_S.gguf) | i1-Q3_K_S | 11.1 | IQ3_XS probably better |
58 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ3_M.gguf) | i1-IQ3_M | 11.6 | |
59 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q3_K_M.gguf) | i1-Q3_K_M | 12.3 | IQ3_S probably better |
60 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q3_K_L.gguf) | i1-Q3_K_L | 13.4 | IQ3_M probably better |
61 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-IQ4_XS.gguf) | i1-IQ4_XS | 13.6 | |
62 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q4_0.gguf) | i1-Q4_0 | 14.4 | fast, low quality |
63 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q4_K_S.gguf) | i1-Q4_K_S | 14.4 | optimal size/speed/quality |
64 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q4_K_M.gguf) | i1-Q4_K_M | 15.3 | fast, recommended |
65 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q4_1.gguf) | i1-Q4_1 | 15.8 | |
66 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q5_K_S.gguf) | i1-Q5_K_S | 17.4 | |
67 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q5_K_M.gguf) | i1-Q5_K_M | 17.8 | |
68 | [GGUF](https://huggingface.co/mradermacher/Delphi-25B-SimpleRL-Math-i1-GGUF/resolve/main/Delphi-25B-SimpleRL-Math.i1-Q6_K.gguf) | i1-Q6_K | 20.6 | practically like static Q6_K |
69
70 Here is a handy graph by ikawrakow comparing some lower-quality quant
71 types (lower is better):
72
73 ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
74
75 And here are Artefact2's thoughts on the matter:
76 https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
77
78 ## FAQ / Model Request
79
80 See https://huggingface.co/mradermacher/model_requests for some answers to
81 questions you might have and/or if you want some other model quantized.
82
83 ## Thanks
84
85 I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
86 me use its servers and providing upgrades to my workstation to enable
87 this work in my free time. Additional thanks to [@nicoboss](https://huggingface.co/nicoboss) for giving me access to his private supercomputer, enabling me to provide many more imatrix quants, at much higher quality, than I would otherwise be able to.
88
89 <!-- end -->
90