mradermacher commited on
Commit
d5c2b71
1 Parent(s): f2ce8f3

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +100 -0
README.md CHANGED
@@ -1,6 +1,106 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  <!-- ### quantize_version: 2 -->
2
  <!-- ### output_tensor_quantised: 1 -->
3
  <!-- ### convert_type: hf -->
4
  <!-- ### vocab_type: -->
5
  <!-- ### tags: -->
6
  static quants of https://huggingface.co/Replete-AI/Replete-Coder-Llama3-8B
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Replete-AI/Replete-Coder-Llama3-8B
3
+ datasets:
4
+ - Replete-AI/code_bagel_hermes-2.5
5
+ - Replete-AI/code_bagel
6
+ - Replete-AI/OpenHermes-2.5-Uncensored
7
+ - teknium/OpenHermes-2.5
8
+ - layoric/tiny-codes-alpaca
9
+ - glaiveai/glaive-code-assistant-v3
10
+ - ajibawa-2023/Code-290k-ShareGPT
11
+ - TIGER-Lab/MathInstruct
12
+ - chargoddard/commitpack-ft-instruct-rated
13
+ - iamturun/code_instructions_120k_alpaca
14
+ - ise-uiuc/Magicoder-Evol-Instruct-110K
15
+ - cognitivecomputations/dolphin-coder
16
+ - nickrosh/Evol-Instruct-Code-80k-v1
17
+ - coseal/CodeUltraFeedback_binarized
18
+ - glaiveai/glaive-function-calling-v2
19
+ - CyberNative/Code_Vulnerability_Security_DPO
20
+ - jondurbin/airoboros-2.2
21
+ - camel-ai
22
+ - lmsys/lmsys-chat-1m
23
+ - CollectiveCognition/chats-data-2023-09-22
24
+ - CoT-Alpaca-GPT4
25
+ - WizardLM/WizardLM_evol_instruct_70k
26
+ - WizardLM/WizardLM_evol_instruct_V2_196k
27
+ - teknium/GPT4-LLM-Cleaned
28
+ - GPTeacher
29
+ - OpenGPT
30
+ - meta-math/MetaMathQA
31
+ - Open-Orca/SlimOrca
32
+ - garage-bAInd/Open-Platypus
33
+ - anon8231489123/ShareGPT_Vicuna_unfiltered
34
+ - Unnatural-Instructions-GPT4
35
+ language:
36
+ - en
37
+ library_name: transformers
38
+ license: other
39
+ license_link: https://llama.meta.com/llama3/license/
40
+ license_name: llama-3
41
+ quantized_by: mradermacher
42
+ tags:
43
+ - text-generation-inference
44
+ - transformers
45
+ - unsloth
46
+ - llama
47
+ ---
48
+ ## About
49
+
50
  <!-- ### quantize_version: 2 -->
51
  <!-- ### output_tensor_quantised: 1 -->
52
  <!-- ### convert_type: hf -->
53
  <!-- ### vocab_type: -->
54
  <!-- ### tags: -->
55
  static quants of https://huggingface.co/Replete-AI/Replete-Coder-Llama3-8B
56
+
57
+ <!-- provided-files -->
58
+ weighted/imatrix quants seem not to be available (by me) at this time. If they do not show up a week or so after the static ones, I have probably not planned for them. Feel free to request them by opening a Community Discussion.
59
+ ## Usage
60
+
61
+ If you are unsure how to use GGUF files, refer to one of [TheBloke's
62
+ READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
63
+ more details, including on how to concatenate multi-part files.
64
+
65
+ ## Provided Quants
66
+
67
+ (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
68
+
69
+ | Link | Type | Size/GB | Notes |
70
+ |:-----|:-----|--------:|:------|
71
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.Q2_K.gguf) | Q2_K | 3.3 | |
72
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.IQ3_XS.gguf) | IQ3_XS | 3.6 | |
73
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.Q3_K_S.gguf) | Q3_K_S | 3.8 | |
74
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.IQ3_S.gguf) | IQ3_S | 3.8 | beats Q3_K* |
75
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.IQ3_M.gguf) | IQ3_M | 3.9 | |
76
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.Q3_K_M.gguf) | Q3_K_M | 4.1 | lower quality |
77
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.Q3_K_L.gguf) | Q3_K_L | 4.4 | |
78
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.IQ4_XS.gguf) | IQ4_XS | 4.6 | |
79
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.Q4_K_S.gguf) | Q4_K_S | 4.8 | fast, recommended |
80
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.Q4_K_M.gguf) | Q4_K_M | 5.0 | fast, recommended |
81
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.Q5_K_S.gguf) | Q5_K_S | 5.7 | |
82
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.Q5_K_M.gguf) | Q5_K_M | 5.8 | |
83
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.Q6_K.gguf) | Q6_K | 6.7 | very good quality |
84
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.Q8_0.gguf) | Q8_0 | 8.6 | fast, best quality |
85
+ | [GGUF](https://huggingface.co/mradermacher/Replete-Coder-Llama3-8B-GGUF/resolve/main/Replete-Coder-Llama3-8B.f16.gguf) | f16 | 16.2 | 16 bpw, overkill |
86
+
87
+ Here is a handy graph by ikawrakow comparing some lower-quality quant
88
+ types (lower is better):
89
+
90
+ ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
91
+
92
+ And here are Artefact2's thoughts on the matter:
93
+ https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
94
+
95
+ ## FAQ / Model Request
96
+
97
+ See https://huggingface.co/mradermacher/model_requests for some answers to
98
+ questions you might have and/or if you want some other model quantized.
99
+
100
+ ## Thanks
101
+
102
+ I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
103
+ me use its servers and providing upgrades to my workstation to enable
104
+ this work in my free time.
105
+
106
+ <!-- end -->