mradermacher
commited on
Commit
•
e0a5d2a
1
Parent(s):
821719c
auto-patch README.md
Browse files
README.md
CHANGED
@@ -1,6 +1,117 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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/darkcloudai/huskylm-2.5-8b
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model: darkcloudai/huskylm-2.5-8b
|
3 |
+
datasets:
|
4 |
+
- darkcloudai-smallmodel-frontieredition
|
5 |
+
- darkcloudai-webdriver-redditcrawl-2023
|
6 |
+
- darkcloudai-unalignment-truthfulness
|
7 |
+
- darkcloudai-generaldpo
|
8 |
+
- ai2_arc
|
9 |
+
- allenai/ultrafeedback_binarized_cleaned
|
10 |
+
- argilla/distilabel-intel-orca-dpo-pairs
|
11 |
+
- jondurbin/airoboros-3.2
|
12 |
+
- codeparrot/apps
|
13 |
+
- facebook/belebele
|
14 |
+
- bluemoon-fandom-1-1-rp-cleaned
|
15 |
+
- boolq
|
16 |
+
- camel-ai/biology
|
17 |
+
- camel-ai/chemistry
|
18 |
+
- camel-ai/math
|
19 |
+
- camel-ai/physics
|
20 |
+
- jondurbin/contextual-dpo-v0.1
|
21 |
+
- jondurbin/gutenberg-dpo-v0.1
|
22 |
+
- jondurbin/py-dpo-v0.1
|
23 |
+
- jondurbin/truthy-dpo-v0.1
|
24 |
+
- LDJnr/Capybara
|
25 |
+
- jondurbin/cinematika-v0.1
|
26 |
+
- WizardLM/WizardLM_evol_instruct_70k
|
27 |
+
- glaiveai/glaive-function-calling-v2
|
28 |
+
- jondurbin/gutenberg-dpo-v0.1
|
29 |
+
- grimulkan/LimaRP-augmented
|
30 |
+
- lmsys/lmsys-chat-1m
|
31 |
+
- ParisNeo/lollms_aware_dataset
|
32 |
+
- TIGER-Lab/MathInstruct
|
33 |
+
- Muennighoff/natural-instructions
|
34 |
+
- openbookqa
|
35 |
+
- kingbri/PIPPA-shareGPT
|
36 |
+
- piqa
|
37 |
+
- Vezora/Tested-22k-Python-Alpaca
|
38 |
+
- ropes
|
39 |
+
- cakiki/rosetta-code
|
40 |
+
- Open-Orca/SlimOrca
|
41 |
+
- b-mc2/sql-create-context
|
42 |
+
- squad_v2
|
43 |
+
- mattpscott/airoboros-summarization
|
44 |
+
- migtissera/Synthia-v1.3
|
45 |
+
- unalignment/toxic-dpo-v0.2
|
46 |
+
- WhiteRabbitNeo/WRN-Chapter-1
|
47 |
+
- WhiteRabbitNeo/WRN-Chapter-2
|
48 |
+
- winogrande
|
49 |
+
language:
|
50 |
+
- en
|
51 |
+
library_name: transformers
|
52 |
+
license: llama3
|
53 |
+
quantized_by: mradermacher
|
54 |
+
tags:
|
55 |
+
- llama-3
|
56 |
+
- huskylm
|
57 |
+
- darkcloudai
|
58 |
+
---
|
59 |
+
## About
|
60 |
+
|
61 |
<!-- ### quantize_version: 2 -->
|
62 |
<!-- ### output_tensor_quantised: 1 -->
|
63 |
<!-- ### convert_type: hf -->
|
64 |
<!-- ### vocab_type: -->
|
65 |
<!-- ### tags: -->
|
66 |
static quants of https://huggingface.co/darkcloudai/huskylm-2.5-8b
|
67 |
+
|
68 |
+
<!-- provided-files -->
|
69 |
+
weighted/imatrix quants are available at https://huggingface.co/mradermacher/huskylm-2.5-8b-i1-GGUF
|
70 |
+
## Usage
|
71 |
+
|
72 |
+
If you are unsure how to use GGUF files, refer to one of [TheBloke's
|
73 |
+
READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
|
74 |
+
more details, including on how to concatenate multi-part files.
|
75 |
+
|
76 |
+
## Provided Quants
|
77 |
+
|
78 |
+
(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
|
79 |
+
|
80 |
+
| Link | Type | Size/GB | Notes |
|
81 |
+
|:-----|:-----|--------:|:------|
|
82 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.Q2_K.gguf) | Q2_K | 3.3 | |
|
83 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.IQ3_XS.gguf) | IQ3_XS | 3.6 | |
|
84 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.Q3_K_S.gguf) | Q3_K_S | 3.8 | |
|
85 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.IQ3_S.gguf) | IQ3_S | 3.8 | beats Q3_K* |
|
86 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.IQ3_M.gguf) | IQ3_M | 3.9 | |
|
87 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.Q3_K_M.gguf) | Q3_K_M | 4.1 | lower quality |
|
88 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.Q3_K_L.gguf) | Q3_K_L | 4.4 | |
|
89 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.IQ4_XS.gguf) | IQ4_XS | 4.6 | |
|
90 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.Q4_K_S.gguf) | Q4_K_S | 4.8 | fast, recommended |
|
91 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.Q4_K_M.gguf) | Q4_K_M | 5.0 | fast, recommended |
|
92 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.Q5_K_S.gguf) | Q5_K_S | 5.7 | |
|
93 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.Q5_K_M.gguf) | Q5_K_M | 5.8 | |
|
94 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.Q6_K.gguf) | Q6_K | 6.7 | very good quality |
|
95 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.Q8_0.gguf) | Q8_0 | 8.6 | fast, best quality |
|
96 |
+
| [GGUF](https://huggingface.co/mradermacher/huskylm-2.5-8b-GGUF/resolve/main/huskylm-2.5-8b.f16.gguf) | f16 | 16.2 | 16 bpw, overkill |
|
97 |
+
|
98 |
+
Here is a handy graph by ikawrakow comparing some lower-quality quant
|
99 |
+
types (lower is better):
|
100 |
+
|
101 |
+
![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
|
102 |
+
|
103 |
+
And here are Artefact2's thoughts on the matter:
|
104 |
+
https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
|
105 |
+
|
106 |
+
## FAQ / Model Request
|
107 |
+
|
108 |
+
See https://huggingface.co/mradermacher/model_requests for some answers to
|
109 |
+
questions you might have and/or if you want some other model quantized.
|
110 |
+
|
111 |
+
## Thanks
|
112 |
+
|
113 |
+
I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
|
114 |
+
me use its servers and providing upgrades to my workstation to enable
|
115 |
+
this work in my free time.
|
116 |
+
|
117 |
+
<!-- end -->
|