bartowski commited on
Commit
350c444
1 Parent(s): 7d0b279

Llamacpp quants

Browse files
.gitattributes CHANGED
@@ -33,3 +33,22 @@ 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
+ LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_M.gguf filter=lfs diff=lfs merge=lfs -text
37
+ LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_S.gguf filter=lfs diff=lfs merge=lfs -text
38
+ LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_XS.gguf filter=lfs diff=lfs merge=lfs -text
39
+ LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_M.gguf filter=lfs diff=lfs merge=lfs -text
40
+ LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_XS.gguf filter=lfs diff=lfs merge=lfs -text
41
+ LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_XXS.gguf filter=lfs diff=lfs merge=lfs -text
42
+ LocalAI-Llama3-8b-Function-Call-v0.2-IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text
43
+ LocalAI-Llama3-8b-Function-Call-v0.2-Q2_K.gguf filter=lfs diff=lfs merge=lfs -text
44
+ LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text
45
+ LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text
46
+ LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text
47
+ LocalAI-Llama3-8b-Function-Call-v0.2-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
48
+ LocalAI-Llama3-8b-Function-Call-v0.2-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text
49
+ LocalAI-Llama3-8b-Function-Call-v0.2-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
50
+ LocalAI-Llama3-8b-Function-Call-v0.2-Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text
51
+ LocalAI-Llama3-8b-Function-Call-v0.2-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text
52
+ LocalAI-Llama3-8b-Function-Call-v0.2-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
53
+ LocalAI-Llama3-8b-Function-Call-v0.2-f32.gguf filter=lfs diff=lfs merge=lfs -text
54
+ LocalAI-Llama3-8b-Function-Call-v0.2.imatrix filter=lfs diff=lfs merge=lfs -text
LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba452ffac62f8e8d6ebba5b1e09a5df23805ded23566f4a2ff19c2de6881b0a1
3
+ size 2948281216
LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03ebee9f4c7d5f6266bf7c420dab774501f291ce095e85a22309323b31d719a1
3
+ size 2758488960
LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_XS.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a48a524cb8897de0f31a921a8fd93cd1703af98a9488c2cbab91e498fc92c2a
3
+ size 2605781888
LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0bc492d572a1b9d62ea3e3ec225bad6c4a7525e7af2bfa390c51270cfba062f1
3
+ size 3784823680
LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_XS.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:efa02c89644577d576bfb8879bee4d462099a240e918938d43b305764b7bce3b
3
+ size 3518747520
LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_XXS.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b3694dea3c47b3bd94a8cee2b059255d683e7c97b289dc78ef9ca7d6a8c3236
3
+ size 3274912640
LocalAI-Llama3-8b-Function-Call-v0.2-IQ4_XS.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ac1617406443df16f7ef56de9cce0786eb16bd55e4090f5b8dcda5b0ca9da89
3
+ size 4447662976
LocalAI-Llama3-8b-Function-Call-v0.2-Q2_K.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fdf2b5dbb7341114d89453865afae03d0a7d214a1548d6bb755fc5912e2e9f2c
3
+ size 3179131776
LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_L.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c618f27ab8a24fd402fb5575a688c2afbd152adef662ecace261d5c14745c95
3
+ size 4321956736
LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cbb8bc9c22bbf9f741ae2db67c45e47da40421bb8c0fba6953406dfa088008d1
3
+ size 4018918272
LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:49e6d36ca059dec584f88a5881ac60dd2ed40a5ea49f8de4d804a29cbcb70448
3
+ size 3664499584
LocalAI-Llama3-8b-Function-Call-v0.2-Q4_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0488194ae0c5cfa6d83bed52704825328fc8900f453daafeca797f47ff68e694
3
+ size 4920734592
LocalAI-Llama3-8b-Function-Call-v0.2-Q4_K_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d75d834b67d3b4b60a6fd6f2e431deca08760c11497384defe96db5db7f37e59
3
+ size 4692669312
LocalAI-Llama3-8b-Function-Call-v0.2-Q5_K_M.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b96d2f295208d81c22224e942ce79623da203017e7b76c2c4fd84b335b7626d8
3
+ size 5732987776
LocalAI-Llama3-8b-Function-Call-v0.2-Q5_K_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:647333e082ac05d41bbd5522035e6b2b803970a2e8aabfe5dff5a82d1252c800
3
+ size 5599294336
LocalAI-Llama3-8b-Function-Call-v0.2-Q6_K.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1492c26afe30565a5bc149317fd5f915ddc52512e71738c58880f9c92fbbb863
3
+ size 6596006784
LocalAI-Llama3-8b-Function-Call-v0.2-Q8_0.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e4fef98375fd455fead68fd6dbb42c4b4099db09121914141843c7597c3e52b
3
+ size 8540771200
LocalAI-Llama3-8b-Function-Call-v0.2-f32.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4a81b920030a0e9046a2f2c67783b66acd669de484cc4a678fe6fd88c3a60a8
3
+ size 32128881216
LocalAI-Llama3-8b-Function-Call-v0.2.imatrix ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f170a946b11a40a14d26dd907b4aa2fa83630da289b9ef2271b5f4786aa4add1
3
+ size 4988171
README.md ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: llama3
3
+ tags:
4
+ - LocalAI
5
+ quantized_by: bartowski
6
+ pipeline_tag: text-generation
7
+ ---
8
+
9
+ ## Llamacpp imatrix Quantizations of LocalAI-Llama3-8b-Function-Call-v0.2
10
+
11
+ Using <a href="https://github.com/ggerganov/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggerganov/llama.cpp/releases/tag/b3070">b3070</a> for quantization.
12
+
13
+ Original model: https://huggingface.co/mudler/LocalAI-Llama3-8b-Function-Call-v0.2
14
+
15
+ All quants made using imatrix option with dataset from [here](https://gist.github.com/bartowski1182/eb213dccb3571f863da82e99418f81e8)
16
+
17
+ ## Prompt format
18
+
19
+ ```
20
+ <|begin_of_text|><|start_header_id|>system<|end_header_id|>
21
+
22
+ {system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>
23
+
24
+ {prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
25
+
26
+ <|eot_id|>
27
+ ```
28
+
29
+ ## Download a file (not the whole branch) from below:
30
+
31
+ | Filename | Quant type | File Size | Description |
32
+ | -------- | ---------- | --------- | ----------- |
33
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-Q8_0.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-Q8_0.gguf) | Q8_0 | 8.54GB | Extremely high quality, generally unneeded but max available quant. |
34
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-Q6_K.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-Q6_K.gguf) | Q6_K | 6.59GB | Very high quality, near perfect, *recommended*. |
35
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-Q5_K_M.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-Q5_K_M.gguf) | Q5_K_M | 5.73GB | High quality, *recommended*. |
36
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-Q5_K_S.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-Q5_K_S.gguf) | Q5_K_S | 5.59GB | High quality, *recommended*. |
37
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-Q4_K_M.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-Q4_K_M.gguf) | Q4_K_M | 4.92GB | Good quality, uses about 4.83 bits per weight, *recommended*. |
38
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-Q4_K_S.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-Q4_K_S.gguf) | Q4_K_S | 4.69GB | Slightly lower quality with more space savings, *recommended*. |
39
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-IQ4_XS.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-IQ4_XS.gguf) | IQ4_XS | 4.44GB | Decent quality, smaller than Q4_K_S with similar performance, *recommended*. |
40
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_L.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_L.gguf) | Q3_K_L | 4.32GB | Lower quality but usable, good for low RAM availability. |
41
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_M.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_M.gguf) | Q3_K_M | 4.01GB | Even lower quality. |
42
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_M.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_M.gguf) | IQ3_M | 3.78GB | Medium-low quality, new method with decent performance comparable to Q3_K_M. |
43
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_S.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-Q3_K_S.gguf) | Q3_K_S | 3.66GB | Low quality, not recommended. |
44
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_XS.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_XS.gguf) | IQ3_XS | 3.51GB | Lower quality, new method with decent performance, slightly better than Q3_K_S. |
45
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_XXS.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-IQ3_XXS.gguf) | IQ3_XXS | 3.27GB | Lower quality, new method with decent performance, comparable to Q3 quants. |
46
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-Q2_K.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-Q2_K.gguf) | Q2_K | 3.17GB | Very low quality but surprisingly usable. |
47
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_M.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_M.gguf) | IQ2_M | 2.94GB | Very low quality, uses SOTA techniques to also be surprisingly usable. |
48
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_S.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_S.gguf) | IQ2_S | 2.75GB | Very low quality, uses SOTA techniques to be usable. |
49
+ | [LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_XS.gguf](https://huggingface.co/bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF/blob/main/LocalAI-Llama3-8b-Function-Call-v0.2-IQ2_XS.gguf) | IQ2_XS | 2.60GB | Very low quality, uses SOTA techniques to be usable. |
50
+
51
+ ## Downloading using huggingface-cli
52
+
53
+ First, make sure you have hugginface-cli installed:
54
+
55
+ ```
56
+ pip install -U "huggingface_hub[cli]"
57
+ ```
58
+
59
+ Then, you can target the specific file you want:
60
+
61
+ ```
62
+ huggingface-cli download bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF --include "LocalAI-Llama3-8b-Function-Call-v0.2-Q4_K_M.gguf" --local-dir ./
63
+ ```
64
+
65
+ 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:
66
+
67
+ ```
68
+ huggingface-cli download bartowski/LocalAI-Llama3-8b-Function-Call-v0.2-GGUF --include "LocalAI-Llama3-8b-Function-Call-v0.2-Q8_0.gguf/*" --local-dir LocalAI-Llama3-8b-Function-Call-v0.2-Q8_0
69
+ ```
70
+
71
+ You can either specify a new local-dir (LocalAI-Llama3-8b-Function-Call-v0.2-Q8_0) or download them all in place (./)
72
+
73
+ ## Which file should I choose?
74
+
75
+ A great write up with charts showing various performances is provided by Artefact2 [here](https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9)
76
+
77
+ 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.
78
+
79
+ 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.
80
+
81
+ 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.
82
+
83
+ Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'.
84
+
85
+ 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.
86
+
87
+ If you want to get more into the weeds, you can check out this extremely useful feature chart:
88
+
89
+ [llama.cpp feature matrix](https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix)
90
+
91
+ 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.
92
+
93
+ 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.
94
+
95
+ 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.
96
+
97
+ Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski