Hugging Face's logo

legraphista
/
Hermes-3-Llama-3.1-8B-IMat-GGUF

Text Generation
GGUF
English
Llama-3
instruct
finetune
chatml
gpt4
synthetic data
distillation
function calling
json mode
axolotl
roleplaying
chat
quantized
GGUF
quantization
imat
imatrix
static
16bit
8bit
6bit
5bit
4bit
3bit
2bit
1bit
conversational
Model card Files Community
legraphista commited on
Commit
0f5eb2f
•
1 Parent(s): d5abdce

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +177 -0
README.md ADDED
@@ -0,0 +1,177 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: NousResearch/Hermes-3-Llama-3.1-8B
3
+ inference: false
4
+ language:
5
+ - en
6
+ library_name: gguf
7
+ license: llama3
8
+ model-index:
9
+ - name: Hermes-3-Llama-3.1-70B
10
+ results: []
11
+ pipeline_tag: text-generation
12
+ quantized_by: legraphista
13
+ tags:
14
+ - Llama-3
15
+ - instruct
16
+ - finetune
17
+ - chatml
18
+ - gpt4
19
+ - synthetic data
20
+ - distillation
21
+ - function calling
22
+ - json mode
23
+ - axolotl
24
+ - roleplaying
25
+ - chat
26
+ - quantized
27
+ - GGUF
28
+ - quantization
29
+ - imat
30
+ - imatrix
31
+ - static
32
+ - 16bit
33
+ - 8bit
34
+ - 6bit
35
+ - 5bit
36
+ - 4bit
37
+ - 3bit
38
+ - 2bit
39
+ - 1bit
40
+ ---
41
+
42
+ # Hermes-3-Llama-3.1-8B-IMat-GGUF
43
+ _Llama.cpp imatrix quantization of NousResearch/Hermes-3-Llama-3.1-8B_
44
+
45
+ Original Model: [NousResearch/Hermes-3-Llama-3.1-8B](https://huggingface.co/NousResearch/Hermes-3-Llama-3.1-8B)
46
+ Original dtype: `BF16` (`bfloat16`)
47
+ Quantized by: llama.cpp [b3592](https://github.com/ggerganov/llama.cpp/releases/tag/b3592)
48
+ IMatrix dataset: [here](https://gist.githubusercontent.com/bartowski1182/eb213dccb3571f863da82e99418f81e8/raw/b2869d80f5c16fd7082594248e80144677736635/calibration_datav3.txt)
49
+
50
+ - [Files](#files)
51
+ - [IMatrix](#imatrix)
52
+ - [Common Quants](#common-quants)
53
+ - [All Quants](#all-quants)
54
+ - [Downloading using huggingface-cli](#downloading-using-huggingface-cli)
55
+ - [Inference](#inference)
56
+ - [Simple chat template](#simple-chat-template)
57
+ - [Chat template with system prompt](#chat-template-with-system-prompt)
58
+ - [Llama.cpp](#llama-cpp)
59
+ - [FAQ](#faq)
60
+ - [Why is the IMatrix not applied everywhere?](#why-is-the-imatrix-not-applied-everywhere)
61
+ - [How do I merge a split GGUF?](#how-do-i-merge-a-split-gguf)
62
+
63
+ ---
64
+
65
+ ## Files
66
+
67
+ ### IMatrix
68
+ Status: ⏳ Processing
69
+ Link: [here](https://huggingface.co/legraphista/Hermes-3-Llama-3.1-8B-IMat-GGUF/blob/main/imatrix.dat)
70
+
71
+ ### Common Quants
72
+ | Filename | Quant type | File Size | Status | Uses IMatrix | Is Split |
73
+ | -------- | ---------- | --------- | ------ | ------------ | -------- |
74
+ | Hermes-3-Llama-3.1-8B.Q8_0 | Q8_0 | - | ⏳ Processing | ⚪ Static | -
75
+ | Hermes-3-Llama-3.1-8B.Q6_K | Q6_K | - | ⏳ Processing | ⚪ Static | -
76
+ | Hermes-3-Llama-3.1-8B.Q4_K | Q4_K | - | ⏳ Processing | 🟢 IMatrix | -
77
+ | Hermes-3-Llama-3.1-8B.Q3_K | Q3_K | - | ⏳ Processing | 🟢 IMatrix | -
78
+ | Hermes-3-Llama-3.1-8B.Q2_K | Q2_K | - | ⏳ Processing | 🟢 IMatrix | -
79
+
80
+
81
+ ### All Quants
82
+ | Filename | Quant type | File Size | Status | Uses IMatrix | Is Split |
83
+ | -------- | ---------- | --------- | ------ | ------------ | -------- |
84
+ | Hermes-3-Llama-3.1-8B.BF16 | BF16 | - | ⏳ Processing | ⚪ Static | -
85
+ | Hermes-3-Llama-3.1-8B.FP16 | F16 | - | ⏳ Processing | ⚪ Static | -
86
+ | Hermes-3-Llama-3.1-8B.Q8_0 | Q8_0 | - | ⏳ Processing | ⚪ Static | -
87
+ | Hermes-3-Llama-3.1-8B.Q6_K | Q6_K | - | ⏳ Processing | ⚪ Static | -
88
+ | Hermes-3-Llama-3.1-8B.Q5_K | Q5_K | - | ⏳ Processing | ⚪ Static | -
89
+ | Hermes-3-Llama-3.1-8B.Q5_K_S | Q5_K_S | - | ⏳ Processing | ⚪ Static | -
90
+ | Hermes-3-Llama-3.1-8B.Q4_K | Q4_K | - | ⏳ Processing | 🟢 IMatrix | -
91
+ | Hermes-3-Llama-3.1-8B.Q4_K_S | Q4_K_S | - | ⏳ Processing | 🟢 IMatrix | -
92
+ | Hermes-3-Llama-3.1-8B.IQ4_NL | IQ4_NL | - | ⏳ Processing | 🟢 IMatrix | -
93
+ | Hermes-3-Llama-3.1-8B.IQ4_XS | IQ4_XS | - | ⏳ Processing | 🟢 IMatrix | -
94
+ | Hermes-3-Llama-3.1-8B.Q3_K | Q3_K | - | ⏳ Processing | 🟢 IMatrix | -
95
+ | Hermes-3-Llama-3.1-8B.Q3_K_L | Q3_K_L | - | ⏳ Processing | 🟢 IMatrix | -
96
+ | Hermes-3-Llama-3.1-8B.Q3_K_S | Q3_K_S | - | ⏳ Processing | 🟢 IMatrix | -
97
+ | Hermes-3-Llama-3.1-8B.IQ3_M | IQ3_M | - | ⏳ Processing | 🟢 IMatrix | -
98
+ | Hermes-3-Llama-3.1-8B.IQ3_S | IQ3_S | - | ⏳ Processing | 🟢 IMatrix | -
99
+ | Hermes-3-Llama-3.1-8B.IQ3_XS | IQ3_XS | - | ⏳ Processing | 🟢 IMatrix | -
100
+ | Hermes-3-Llama-3.1-8B.IQ3_XXS | IQ3_XXS | - | ⏳ Processing | 🟢 IMatrix | -
101
+ | Hermes-3-Llama-3.1-8B.Q2_K | Q2_K | - | ⏳ Processing | 🟢 IMatrix | -
102
+ | Hermes-3-Llama-3.1-8B.Q2_K_S | Q2_K_S | - | ⏳ Processing | 🟢 IMatrix | -
103
+ | Hermes-3-Llama-3.1-8B.IQ2_M | IQ2_M | - | ⏳ Processing | 🟢 IMatrix | -
104
+ | Hermes-3-Llama-3.1-8B.IQ2_S | IQ2_S | - | ⏳ Processing | 🟢 IMatrix | -
105
+ | Hermes-3-Llama-3.1-8B.IQ2_XS | IQ2_XS | - | ⏳ Processing | 🟢 IMatrix | -
106
+ | Hermes-3-Llama-3.1-8B.IQ2_XXS | IQ2_XXS | - | ⏳ Processing | 🟢 IMatrix | -
107
+ | Hermes-3-Llama-3.1-8B.IQ1_M | IQ1_M | - | ⏳ Processing | 🟢 IMatrix | -
108
+ | Hermes-3-Llama-3.1-8B.IQ1_S | IQ1_S | - | ⏳ Processing | 🟢 IMatrix | -
109
+
110
+
111
+ ## Downloading using huggingface-cli
112
+ If you do not have hugginface-cli installed:
113
+ ```
114
+ pip install -U "huggingface_hub[cli]"
115
+ ```
116
+ Download the specific file you want:
117
+ ```
118
+ huggingface-cli download legraphista/Hermes-3-Llama-3.1-8B-IMat-GGUF --include "Hermes-3-Llama-3.1-8B.Q8_0.gguf" --local-dir ./
119
+ ```
120
+ If the model file is big, it has been split into multiple files. In order to download them all to a local folder, run:
121
+ ```
122
+ huggingface-cli download legraphista/Hermes-3-Llama-3.1-8B-IMat-GGUF --include "Hermes-3-Llama-3.1-8B.Q8_0/*" --local-dir ./
123
+ # see FAQ for merging GGUF's
124
+ ```
125
+
126
+ ---
127
+
128
+ ## Inference
129
+
130
+ ### Simple chat template
131
+ ```
132
+ <|begin_of_text|><|im_start|>user
133
+ {user_prompt}<|im_end|>
134
+ <|im_start|>assistant
135
+ {assistant_response}<|im_end|>
136
+ <|im_start|>user
137
+ {next_user_prompt}<|im_end|>
138
+
139
+ ```
140
+
141
+ ### Chat template with system prompt
142
+ ```
143
+ <|begin_of_text|><|im_start|>system
144
+ {system_prompt}<|im_end|>
145
+ <|im_start|>user
146
+ {user_prompt}<|im_end|>
147
+ <|im_start|>assistant
148
+ {assistant_response}<|im_end|>
149
+ <|im_start|>user
150
+ {next_user_prompt}<|im_end|>
151
+
152
+ ```
153
+
154
+ ### Llama.cpp
155
+ ```
156
+ llama.cpp/main -m Hermes-3-Llama-3.1-8B.Q8_0.gguf --color -i -p "prompt here (according to the chat template)"
157
+ ```
158
+
159
+ ---
160
+
161
+ ## FAQ
162
+
163
+ ### Why is the IMatrix not applied everywhere?
164
+ According to [this investigation](https://www.reddit.com/r/LocalLLaMA/comments/1993iro/ggufs_quants_can_punch_above_their_weights_now/), it appears that lower quantizations are the only ones that benefit from the imatrix input (as per hellaswag results).
165
+
166
+ ### How do I merge a split GGUF?
167
+ 1. Make sure you have `gguf-split` available
168
+ - To get hold of `gguf-split`, navigate to https://github.com/ggerganov/llama.cpp/releases
169
+ - Download the appropriate zip for your system from the latest release
170
+ - Unzip the archive and you should be able to find `gguf-split`
171
+ 2. Locate your GGUF chunks folder (ex: `Hermes-3-Llama-3.1-8B.Q8_0`)
172
+ 3. Run `gguf-split --merge Hermes-3-Llama-3.1-8B.Q8_0/Hermes-3-Llama-3.1-8B.Q8_0-00001-of-XXXXX.gguf Hermes-3-Llama-3.1-8B.Q8_0.gguf`
173
+ - Make sure to point `gguf-split` to the first chunk of the split.
174
+
175
+ ---
176
+
177
+ Got a suggestion? Ping me [@legraphista](https://x.com/legraphista)!