bartowski commited on
Commit
c0b73d5
1 Parent(s): 45b3e36

Llamacpp quants

Browse files
.gitattributes CHANGED
@@ -33,3 +33,26 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ Samantha-Qwen-2-7B-IQ2_M.gguf filter=lfs diff=lfs merge=lfs -text
37
+ Samantha-Qwen-2-7B-IQ2_S.gguf filter=lfs diff=lfs merge=lfs -text
38
+ Samantha-Qwen-2-7B-IQ2_XS.gguf filter=lfs diff=lfs merge=lfs -text
39
+ Samantha-Qwen-2-7B-IQ3_M.gguf filter=lfs diff=lfs merge=lfs -text
40
+ Samantha-Qwen-2-7B-IQ3_XS.gguf filter=lfs diff=lfs merge=lfs -text
41
+ Samantha-Qwen-2-7B-IQ3_XXS.gguf filter=lfs diff=lfs merge=lfs -text
42
+ Samantha-Qwen-2-7B-IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text
43
+ Samantha-Qwen-2-7B-Q2_K.gguf filter=lfs diff=lfs merge=lfs -text
44
+ Samantha-Qwen-2-7B-Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text
45
+ Samantha-Qwen-2-7B-Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text
46
+ Samantha-Qwen-2-7B-Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text
47
+ Samantha-Qwen-2-7B-Q4_K_L.gguf filter=lfs diff=lfs merge=lfs -text
48
+ Samantha-Qwen-2-7B-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
49
+ Samantha-Qwen-2-7B-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text
50
+ Samantha-Qwen-2-7B-Q5_K_L.gguf filter=lfs diff=lfs merge=lfs -text
51
+ Samantha-Qwen-2-7B-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
52
+ Samantha-Qwen-2-7B-Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text
53
+ Samantha-Qwen-2-7B-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text
54
+ Samantha-Qwen-2-7B-Q6_K_L.gguf filter=lfs diff=lfs merge=lfs -text
55
+ Samantha-Qwen-2-7B-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
56
+ Samantha-Qwen-2-7B-Q8_1.gguf filter=lfs diff=lfs merge=lfs -text
57
+ Samantha-Qwen-2-7B-f32.gguf filter=lfs diff=lfs merge=lfs -text
58
+ Samantha-Qwen-2-7B.imatrix filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,104 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Qwen/Qwen2-7B
3
+ datasets:
4
+ - macadeliccc/opus_samantha
5
+ - cognitivecomputations/ultrachat-uncensored
6
+ - teknium/OpenHermes-2.5
7
+ - Sao10K/Claude-3-Opus-Instruct-15K
8
+ license: apache-2.0
9
+ quantized_by: bartowski
10
+ pipeline_tag: text-generation
11
+ ---
12
+
13
+ ## Llamacpp imatrix Quantizations of Samantha-Qwen-2-7B
14
+
15
+ Using <a href="https://github.com/ggerganov/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggerganov/llama.cpp/releases/tag/b3152">b3152</a> for quantization.
16
+
17
+ Original model: https://huggingface.co/macadeliccc/Samantha-Qwen-2-7B
18
+
19
+ All quants made using imatrix option with dataset from [here](https://gist.github.com/bartowski1182/eb213dccb3571f863da82e99418f81e8)
20
+
21
+ ## Prompt format
22
+
23
+ ```
24
+ <|im_start|>system
25
+ {system_prompt}<|im_end|>
26
+ <|im_start|>user
27
+ {prompt}<|im_end|>
28
+ <|im_start|>assistant
29
+
30
+ ```
31
+
32
+ ## Download a file (not the whole branch) from below:
33
+
34
+ | Filename | Quant type | File Size | Description |
35
+ | -------- | ---------- | --------- | ----------- |
36
+ | [Samantha-Qwen-2-7B-Q8_1.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q8_1.gguf) | Q8_1 | 9.12GB | *Experimental*, uses f16 for embed and output weights. Please provide any feedback of differences. Extremely high quality, generally unneeded but max available quant. |
37
+ | [Samantha-Qwen-2-7B-Q8_0.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q8_0.gguf) | Q8_0 | 8.09GB | Extremely high quality, generally unneeded but max available quant. |
38
+ | [Samantha-Qwen-2-7B-Q6_K_L.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF//main/Samantha-Qwen-2-7B-Q6_K_L.gguf) | Q6_K_L | 7.54GB | *Experimental*, uses f16 for embed and output weights. Please provide any feedback of differences. Very high quality, near perfect, *recommended*. |
39
+ | [Samantha-Qwen-2-7B-Q6_K.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q6_K.gguf) | Q6_K | 6.25GB | Very high quality, near perfect, *recommended*. |
40
+ | [Samantha-Qwen-2-7B-Q5_K_L.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q5_K_L.gguf) | Q5_K_L | 6.80GB | High quality, *recommended*. |
41
+ | [Samantha-Qwen-2-7B-Q5_K_M.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q5_K_M.gguf) | Q5_K_M | 5.44GB | High quality, *recommended*. |
42
+ | [Samantha-Qwen-2-7B-Q5_K_S.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q5_K_S.gguf) | Q5_K_S | 5.31GB | High quality, *recommended*. |
43
+ | [Samantha-Qwen-2-7B-Q4_K_L.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q4_K_L.gguf) | Q4_K_L | 6.10GB | *Experimental*, uses f16 for embed and output weights. Please provide any feedback of differences. Good quality, uses about 4.83 bits per weight, *recommended*. |
44
+ | [Samantha-Qwen-2-7B-Q4_K_M.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q4_K_M.gguf) | Q4_K_M | 4.68GB | Good quality, uses about 4.83 bits per weight, *recommended*. |
45
+ | [Samantha-Qwen-2-7B-Q4_K_S.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q4_K_S.gguf) | Q4_K_S | 4.45GB | Slightly lower quality with more space savings, *recommended*. |
46
+ | [Samantha-Qwen-2-7B-IQ4_XS.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-IQ4_XS.gguf) | IQ4_XS | 4.21GB | Decent quality, smaller than Q4_K_S with similar performance, *recommended*. |
47
+ | [Samantha-Qwen-2-7B-Q3_K_L.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q3_K_L.gguf) | Q3_K_L | 4.08GB | Lower quality but usable, good for low RAM availability. |
48
+ | [Samantha-Qwen-2-7B-Q3_K_M.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q3_K_M.gguf) | Q3_K_M | 3.80GB | Even lower quality. |
49
+ | [Samantha-Qwen-2-7B-IQ3_M.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-IQ3_M.gguf) | IQ3_M | 3.57GB | Medium-low quality, new method with decent performance comparable to Q3_K_M. |
50
+ | [Samantha-Qwen-2-7B-Q3_K_S.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q3_K_S.gguf) | Q3_K_S | 3.49GB | Low quality, not recommended. |
51
+ | [Samantha-Qwen-2-7B-IQ3_XS.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-IQ3_XS.gguf) | IQ3_XS | 3.34GB | Lower quality, new method with decent performance, slightly better than Q3_K_S. |
52
+ | [Samantha-Qwen-2-7B-IQ3_XXS.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-IQ3_XXS.gguf) | IQ3_XXS | 3.11GB | Lower quality, new method with decent performance, comparable to Q3 quants. |
53
+ | [Samantha-Qwen-2-7B-Q2_K.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-Q2_K.gguf) | Q2_K | 3.01GB | Very low quality but surprisingly usable. |
54
+ | [Samantha-Qwen-2-7B-IQ2_M.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-IQ2_M.gguf) | IQ2_M | 2.78GB | Very low quality, uses SOTA techniques to also be surprisingly usable. |
55
+ | [Samantha-Qwen-2-7B-IQ2_S.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-IQ2_S.gguf) | IQ2_S | 2.59GB | Very low quality, uses SOTA techniques to be usable. |
56
+ | [Samantha-Qwen-2-7B-IQ2_XS.gguf](https://huggingface.co/bartowski/Samantha-Qwen-2-7B-GGUF/blob/main/Samantha-Qwen-2-7B-IQ2_XS.gguf) | IQ2_XS | 2.46GB | Very low quality, uses SOTA techniques to be usable. |
57
+
58
+ ## Downloading using huggingface-cli
59
+
60
+ First, make sure you have hugginface-cli installed:
61
+
62
+ ```
63
+ pip install -U "huggingface_hub[cli]"
64
+ ```
65
+
66
+ Then, you can target the specific file you want:
67
+
68
+ ```
69
+ huggingface-cli download bartowski/Samantha-Qwen-2-7B-GGUF --include "Samantha-Qwen-2-7B-Q4_K_M.gguf" --local-dir ./
70
+ ```
71
+
72
+ If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:
73
+
74
+ ```
75
+ huggingface-cli download bartowski/Samantha-Qwen-2-7B-GGUF --include "Samantha-Qwen-2-7B-Q8_0.gguf/*" --local-dir Samantha-Qwen-2-7B-Q8_0
76
+ ```
77
+
78
+ You can either specify a new local-dir (Samantha-Qwen-2-7B-Q8_0) or download them all in place (./)
79
+
80
+ ## Which file should I choose?
81
+
82
+ A great write up with charts showing various performances is provided by Artefact2 [here](https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9)
83
+
84
+ The first thing to figure out is how big a model you can run. To do this, you'll need to figure out how much RAM and/or VRAM you have.
85
+
86
+ If you want your model running as FAST as possible, you'll want to fit the whole thing on your GPU's VRAM. Aim for a quant with a file size 1-2GB smaller than your GPU's total VRAM.
87
+
88
+ If you want the absolute maximum quality, add both your system RAM and your GPU's VRAM together, then similarly grab a quant with a file size 1-2GB Smaller than that total.
89
+
90
+ Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'.
91
+
92
+ If you don't want to think too much, grab one of the K-quants. These are in format 'QX_K_X', like Q5_K_M.
93
+
94
+ If you want to get more into the weeds, you can check out this extremely useful feature chart:
95
+
96
+ [llama.cpp feature matrix](https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix)
97
+
98
+ But basically, if you're aiming for below Q4, and you're running cuBLAS (Nvidia) or rocBLAS (AMD), you should look towards the I-quants. These are in format IQX_X, like IQ3_M. These are newer and offer better performance for their size.
99
+
100
+ These I-quants can also be used on CPU and Apple Metal, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide.
101
+
102
+ The I-quants are *not* compatible with Vulcan, which is also AMD, so if you have an AMD card double check if you're using the rocBLAS build or the Vulcan build. At the time of writing this, LM Studio has a preview with ROCm support, and other inference engines have specific builds for ROCm.
103
+
104
+ Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski
Samantha-Qwen-2-7B-IQ2_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d03e13097ea0fca1a5e166b3a4534ad89ed997801430cd71847f99ed68668bf1
3
+ size 2780340000
Samantha-Qwen-2-7B-IQ2_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc0cf7070453fa70fb66ebc5d2bfa6a25c6ac76dae1a03444d6484ff0c83a1f9
3
+ size 2595634976
Samantha-Qwen-2-7B-IQ2_XS.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53db21c29cbf69e170835e5988cbd47a1289d38bd5bfeee4cb6b966334ae79c4
3
+ size 2469019424
Samantha-Qwen-2-7B-IQ3_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b8a3d011705466030ca2b2f81386355e3a5eeed12d1f6cdfa31190d6bbe3f38
3
+ size 3574009632
Samantha-Qwen-2-7B-IQ3_XS.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9266fdb38803740f1d5084d8269bf355789850565b7abb95672b21033117a262
3
+ size 3346253600
Samantha-Qwen-2-7B-IQ3_XXS.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:608ba605fcd63cc6681fdac7a226fc711351b80a43678d7903b857e487f0d33b
3
+ size 3114512160
Samantha-Qwen-2-7B-IQ4_XS.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9221782a155615e46ccda12bbebb01e47ef53d29bf6691041d02028aaa39146d
3
+ size 4218470176
Samantha-Qwen-2-7B-Q2_K.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0da01bee610ce39af3535aebaf42bf2028babf56c6d9f351e27338369c64eb99
3
+ size 3015937824
Samantha-Qwen-2-7B-Q3_K_L.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f288a7bfa1652dfef3e16d5bbacd2c6cef46403c3401263abcc4925b0961da3f
3
+ size 4088456992
Samantha-Qwen-2-7B-Q3_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3e76d12f67af7ef55c5be71122b349ce36b6936b5a5ca511049050f1b1589e3
3
+ size 3808388896
Samantha-Qwen-2-7B-Q3_K_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fdf1ce82cb61d01be5eaa4add0334b8b1c9cc18837785355338d3500bc39760c
3
+ size 3492366112
Samantha-Qwen-2-7B-Q4_K_L.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5cc29f0d8016e3db194be0a2c7e434132b20ad482843ea62fe49b02b47337759
3
+ size 6109431584
Samantha-Qwen-2-7B-Q4_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d1cf1c35a7a46c536a96ba0417d08b9f9e09c24a4e25976f72ad55d4904f6fe
3
+ size 4683071264
Samantha-Qwen-2-7B-Q4_K_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d10ee9e8d97908d46d75087d6b03bda47eeeb7548aec587d2732c6620a67ed3f
3
+ size 4457766688
Samantha-Qwen-2-7B-Q5_K_L.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e8146ba8dbab15485e9cd3dd72367a538fb2a8d3e9df906209dc0dc3778fced
3
+ size 6803064608
Samantha-Qwen-2-7B-Q5_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b202cdb0449571d5550809845d028b73e67b4bb01539066ea69b18ea19bfbb1
3
+ size 5444828960
Samantha-Qwen-2-7B-Q5_K_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2737ba8454109f158025ef18e2394c84d06aa23d00fc675b52ed39f7d36db49f
3
+ size 5315174176
Samantha-Qwen-2-7B-Q6_K.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9879861d6871abb671bcaeb844901728286e2cc1d54003eaae5cc0d00daf64fe
3
+ size 6254196512
Samantha-Qwen-2-7B-Q6_K_L.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:47144eea889f121c3fdb8c6dc075c096b827cfc6c807de695506a42f499d61df
3
+ size 7540049696
Samantha-Qwen-2-7B-Q8_0.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0e3aa6f20990d4cfbab4b1877e74a312b91e639ec2b56f144c45763b63aa281
3
+ size 8098522912
Samantha-Qwen-2-7B-Q8_1.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96f316e7b148e374689f7ac3078bf1b046caded60b009b07482825164ffe82e2
3
+ size 9120392992
Samantha-Qwen-2-7B-f32.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:35160bfec445f6d7f3b1b250770c409a4d48a5de332d79f8ea2861d59a902463
3
+ size 30468417056
Samantha-Qwen-2-7B.imatrix ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d085980126e40db05910e724623d1601a820312db1d1ae6e41dd51d579429f1
3
+ size 4536679