README.md
6.7 KB · 102 lines · markdown Raw
1 ---
2 base_model: NovatasticRoScript/Atomight-V2.1-0.5B-Inference
3 datasets:
4 - openai/gsm8k
5 - openai/openai_humaneval
6 - cais/mmlu
7 - allenai/ai2_arc
8 - bespokelabs/Bespoke-Stratos-17k
9 language:
10 - en
11 library_name: transformers
12 license: mit
13 mradermacher:
14 readme_rev: 1
15 quantized_by: mradermacher
16 tags:
17 - text-generation
18 - causal-lm
19 - grpo
20 - reasoning
21 - reinforcement-learning
22 - mini-llm
23 - text-generation-inference
24 - lm-evaluation-harness
25 - AtomightV2.1
26 ---
27 ## About
28
29 <!-- ### quantize_version: 2 -->
30 <!-- ### output_tensor_quantised: 1 -->
31 <!-- ### convert_type: hf -->
32 <!-- ### vocab_type: -->
33 <!-- ### tags: nicoboss -->
34 <!-- ### 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 -->
35 <!-- ### quants_skip: -->
36 <!-- ### skip_mmproj: -->
37 weighted/imatrix quants of https://huggingface.co/NovatasticRoScript/Atomight-V2.1-0.5B-Inference
38
39 <!-- provided-files -->
40
41 ***For a convenient overview and download list, visit our [model page for this model](https://hf.tst.eu/model#Atomight-V2.1-0.5B-Inference-i1-GGUF).***
42
43 static quants are available at https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-GGUF
44 ## Usage
45
46 If you are unsure how to use GGUF files, refer to one of [TheBloke's
47 READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
48 more details, including on how to concatenate multi-part files.
49
50 ## Provided Quants
51
52 (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
53
54 | Link | Type | Size/GB | Notes |
55 |:-----|:-----|--------:|:------|
56 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.imatrix.gguf) | imatrix | 0.1 | imatrix file (for creating your own quants) |
57 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ1_S.gguf) | i1-IQ1_S | 0.4 | for the desperate |
58 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ1_M.gguf) | i1-IQ1_M | 0.4 | mostly desperate |
59 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 0.4 | |
60 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ2_XS.gguf) | i1-IQ2_XS | 0.4 | |
61 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ2_S.gguf) | i1-IQ2_S | 0.4 | |
62 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ2_M.gguf) | i1-IQ2_M | 0.4 | |
63 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q2_K_S.gguf) | i1-Q2_K_S | 0.4 | very low quality |
64 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 0.4 | lower quality |
65 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q3_K_S.gguf) | i1-Q3_K_S | 0.4 | IQ3_XS probably better |
66 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ3_S.gguf) | i1-IQ3_S | 0.4 | beats Q3_K* |
67 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ3_XS.gguf) | i1-IQ3_XS | 0.4 | |
68 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q2_K.gguf) | i1-Q2_K | 0.4 | IQ3_XXS probably better |
69 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ3_M.gguf) | i1-IQ3_M | 0.4 | |
70 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ4_XS.gguf) | i1-IQ4_XS | 0.4 | |
71 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-IQ4_NL.gguf) | i1-IQ4_NL | 0.5 | prefer IQ4_XS |
72 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q4_0.gguf) | i1-Q4_0 | 0.5 | fast, low quality |
73 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q3_K_M.gguf) | i1-Q3_K_M | 0.5 | IQ3_S probably better |
74 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q3_K_L.gguf) | i1-Q3_K_L | 0.5 | IQ3_M probably better |
75 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q4_1.gguf) | i1-Q4_1 | 0.5 | |
76 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q4_K_S.gguf) | i1-Q4_K_S | 0.5 | optimal size/speed/quality |
77 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q4_K_M.gguf) | i1-Q4_K_M | 0.5 | fast, recommended |
78 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q5_K_S.gguf) | i1-Q5_K_S | 0.5 | |
79 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q5_K_M.gguf) | i1-Q5_K_M | 0.5 | |
80 | [GGUF](https://huggingface.co/mradermacher/Atomight-V2.1-0.5B-Inference-i1-GGUF/resolve/main/Atomight-V2.1-0.5B-Inference.i1-Q6_K.gguf) | i1-Q6_K | 0.6 | practically like static Q6_K |
81
82 Here is a handy graph by ikawrakow comparing some lower-quality quant
83 types (lower is better):
84
85 ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
86
87 And here are Artefact2's thoughts on the matter:
88 https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
89
90 ## FAQ / Model Request
91
92 See https://huggingface.co/mradermacher/model_requests for some answers to
93 questions you might have and/or if you want some other model quantized.
94
95 ## Thanks
96
97 I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
98 me use its servers and providing upgrades to my workstation to enable
99 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.
100
101 <!-- end -->
102