Atmpushout/Llama-3.2-3B-Instruct-uncensored-GGUF
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

初始化项目，由ModelHub XC社区提供模型

Browse Source 
Model: Atmpushout/Llama-3.2-3B-Instruct-uncensored-GGUF
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-06-12 23:39:19 +08:00
commit 5c17831bca
26 changed files with 258 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

59
.gitattributes vendored Normal file
View File

@@ -0,0 +1,59 @@
*.7z filter=lfs diff=lfs merge=lfs -text
*.arrow filter=lfs diff=lfs merge=lfs -text
*.bin filter=lfs diff=lfs merge=lfs -text
*.bz2 filter=lfs diff=lfs merge=lfs -text
*.ckpt filter=lfs diff=lfs merge=lfs -text
*.ftz filter=lfs diff=lfs merge=lfs -text
*.gz filter=lfs diff=lfs merge=lfs -text
*.h5 filter=lfs diff=lfs merge=lfs -text
*.joblib filter=lfs diff=lfs merge=lfs -text
*.lfs.* filter=lfs diff=lfs merge=lfs -text
*.mlmodel filter=lfs diff=lfs merge=lfs -text
*.model filter=lfs diff=lfs merge=lfs -text
*.msgpack filter=lfs diff=lfs merge=lfs -text
*.npy filter=lfs diff=lfs merge=lfs -text
*.npz filter=lfs diff=lfs merge=lfs -text
*.onnx filter=lfs diff=lfs merge=lfs -text
*.ot filter=lfs diff=lfs merge=lfs -text
*.parquet filter=lfs diff=lfs merge=lfs -text
*.pb filter=lfs diff=lfs merge=lfs -text
*.pickle filter=lfs diff=lfs merge=lfs -text
*.pkl filter=lfs diff=lfs merge=lfs -text
*.pt filter=lfs diff=lfs merge=lfs -text
*.pth filter=lfs diff=lfs merge=lfs -text
*.rar filter=lfs diff=lfs merge=lfs -text
*.safetensors filter=lfs diff=lfs merge=lfs -text
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
*.tar.* filter=lfs diff=lfs merge=lfs -text
*.tar filter=lfs diff=lfs merge=lfs -text
*.tflite filter=lfs diff=lfs merge=lfs -text
*.tgz filter=lfs diff=lfs merge=lfs -text
*.wasm filter=lfs diff=lfs merge=lfs -text
*.xz filter=lfs diff=lfs merge=lfs -text
*.zip filter=lfs diff=lfs merge=lfs -text
*.zst filter=lfs diff=lfs merge=lfs -text
*tfevents* filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q6_K_L.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q5_K_L.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q4_K_L.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q4_0_8_8.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q4_0_4_8.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q4_0_4_4.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q4_0.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q3_K_XL.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-IQ3_M.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-IQ3_XS.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q2_K_L.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-Q2_K.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored-f16.gguf filter=lfs diff=lfs merge=lfs -text
Llama-3.2-3B-Instruct-uncensored.imatrix filter=lfs diff=lfs merge=lfs -text

3
Llama-3.2-3B-Instruct-uncensored-IQ3_M.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:6231f0dc4ed24a7256a532d31f2141117d8913dbd72fefbd9e02c449deaa75f5
size 1768966528

3
Llama-3.2-3B-Instruct-uncensored-IQ3_XS.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:85aade6428d61ec5ccbc278db8153b7bea57b8b826146548cd62f8b4f04f72a6
size 1646086528

3
Llama-3.2-3B-Instruct-uncensored-IQ4_XS.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:029cbe615d4975bb79cb9d7032e629b9331db08085c563a81e322b2a4f68d3e7
size 2038423936

3
Llama-3.2-3B-Instruct-uncensored-Q2_K.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:62cb26657b788af2f2f049d253d2fdc32db813ef66fad61ccd890b7ebed47212
size 1493217664

3
Llama-3.2-3B-Instruct-uncensored-Q2_K_L.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:0cafd957f03a22a5ac4569021255fce506a29b57aefd8ef7c3cb7eed0df442cc
size 1877985664

3
Llama-3.2-3B-Instruct-uncensored-Q3_K_L.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:2c1226c90c854c0c4a13e1d284c742de002d7ce6dd593e48fb44b6058f8e3738
size 1984645504

3
Llama-3.2-3B-Instruct-uncensored-Q3_K_M.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:f77cd97f789164744442fe916f1e8f3c6fc531fc45620ae965a4b7d6fad2b207
size 1856457088

3
Llama-3.2-3B-Instruct-uncensored-Q3_K_S.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:8b883b10870ef381f7f51f8d1fae9fa13d9f466eeca8887c34381fd927551e79
size 1712146816

3
Llama-3.2-3B-Instruct-uncensored-Q3_K_XL.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:238d853fcb86a7b4ca59ad2b5ce4d71ce240c81f779b9a8878a347b13e7110e0
size 2329397632

3
Llama-3.2-3B-Instruct-uncensored-Q4_0.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:185f1bbe74422aedcf299929f0869ba1f9bf55841627ea96e1f2e0ff29cf0acf
size 2143535488

3
Llama-3.2-3B-Instruct-uncensored-Q4_0_4_4.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:ed7ec3a830f3dc5799cfa6c176a958fc14d35330bb4b007af670cbdb72adf507
size 2138816896

3
Llama-3.2-3B-Instruct-uncensored-Q4_0_4_8.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:51345d7d7f4a513647990f684e58810593bc1c9d39147f097794162af6560719
size 2138816896

3
Llama-3.2-3B-Instruct-uncensored-Q4_0_8_8.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:0f5f2f775ba30faa9b60bcee19dc320eb2415a89630f8ca71692f75594fee5ea
size 2138816896

3
Llama-3.2-3B-Instruct-uncensored-Q4_K_L.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:840bf09edb7e2278bd4a998e4fc5f8bdc924b77a3538eac3196aa219ae533bbf
size 2533427584

3
Llama-3.2-3B-Instruct-uncensored-Q4_K_M.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:80f532552e3d56e366226f428395de8285a671f2da1d5fd68563741181b77a95
size 2241003904

3
Llama-3.2-3B-Instruct-uncensored-Q4_K_S.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:826fe076e7fbb119720eba745c0de2e2387caed4bc072118eb360f397c6362c7
size 2149826944

3
Llama-3.2-3B-Instruct-uncensored-Q5_K_L.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:e571f0e2cff8447d78a303588615ef6978a06bd2a7aef577de6cb49ac2b81d00
size 2836203904

3
Llama-3.2-3B-Instruct-uncensored-Q5_K_M.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:492bee2f12826e4d5006532c697332a7db041c700adf9d77d82fc664597f0b10
size 2593030528

3
Llama-3.2-3B-Instruct-uncensored-Q5_K_S.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:cddfa4071a773908f8456eb4fdc7f8ea5515ce2f998119854839b8430a9a0d6e
size 2540388736

3
Llama-3.2-3B-Instruct-uncensored-Q6_K.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:0ac5a51887349714283520c2fc7522d170c6938620ecc2915e8ee41cb9c24e18
size 2967058816

3
Llama-3.2-3B-Instruct-uncensored-Q6_K_L.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:30154318cece42d3e8913f0e229228ce96087d73554878137ae9f6e19f15f672
size 3157903744

3
Llama-3.2-3B-Instruct-uncensored-Q8_0.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:4252d3b5b2a63ff31b515585b6f01a089b71c93ab3ac3108163a612b544d5cfb
size 3840526720

3
Llama-3.2-3B-Instruct-uncensored-f16.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:f9c02234111b72388632b4f9f83f2ea510bd871cd5f74f41d26392e6cc49f68d
size 7221692512

3
Llama-3.2-3B-Instruct-uncensored.imatrix Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:5f3c43443c006752baf7f1c898abf1c75c34a8f02f9ce7b8d8048d19021e885b
size 2988390

127
README.md Normal file
View File

@@ -0,0 +1,127 @@
---
base_model: chuanli11/Llama-3.2-3B-Instruct-uncensored
pipeline_tag: text-generation
tags: []
quantized_by: bartowski
---
## Llamacpp imatrix Quantizations of Llama-3.2-3B-Instruct-uncensored
Using <a href="https://github.com/ggerganov/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggerganov/llama.cpp/releases/tag/b3972">b3972</a> for quantization.
Original model: https://huggingface.co/chuanli11/Llama-3.2-3B-Instruct-uncensored
All quants made using imatrix option with dataset from [here](https://gist.github.com/bartowski1182/eb213dccb3571f863da82e99418f81e8)
Run them in [LM Studio](https://lmstudio.ai/)
## Prompt format
```
<|begin_of_text|><|start_header_id|>system<|end_header_id|>
Cutting Knowledge Date: December 2023
Today Date: 25 Oct 2024
{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>
{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
```
## Download a file (not the whole branch) from below:
| Filename | Quant type | File Size | Split | Description |
| -------- | ---------- | --------- | ----- | ----------- |
| [Llama-3.2-3B-Instruct-uncensored-f16.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-f16.gguf) | f16 | 7.22GB | false | Full F16 weights. |
| [Llama-3.2-3B-Instruct-uncensored-Q8_0.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q8_0.gguf) | Q8_0 | 3.84GB | false | Extremely high quality, generally unneeded but max available quant. |
| [Llama-3.2-3B-Instruct-uncensored-Q6_K_L.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q6_K_L.gguf) | Q6_K_L | 3.16GB | false | Uses Q8_0 for embed and output weights. Very high quality, near perfect, *recommended*. |
| [Llama-3.2-3B-Instruct-uncensored-Q6_K.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q6_K.gguf) | Q6_K | 2.97GB | false | Very high quality, near perfect, *recommended*. |
| [Llama-3.2-3B-Instruct-uncensored-Q5_K_L.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q5_K_L.gguf) | Q5_K_L | 2.84GB | false | Uses Q8_0 for embed and output weights. High quality, *recommended*. |
| [Llama-3.2-3B-Instruct-uncensored-Q5_K_M.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q5_K_M.gguf) | Q5_K_M | 2.59GB | false | High quality, *recommended*. |
| [Llama-3.2-3B-Instruct-uncensored-Q5_K_S.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q5_K_S.gguf) | Q5_K_S | 2.54GB | false | High quality, *recommended*. |
| [Llama-3.2-3B-Instruct-uncensored-Q4_K_L.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q4_K_L.gguf) | Q4_K_L | 2.53GB | false | Uses Q8_0 for embed and output weights. Good quality, *recommended*. |
| [Llama-3.2-3B-Instruct-uncensored-Q3_K_XL.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q3_K_XL.gguf) | Q3_K_XL | 2.33GB | false | Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability. |
| [Llama-3.2-3B-Instruct-uncensored-Q4_K_M.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q4_K_M.gguf) | Q4_K_M | 2.24GB | false | Good quality, default size for must use cases, *recommended*. |
| [Llama-3.2-3B-Instruct-uncensored-Q4_K_S.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q4_K_S.gguf) | Q4_K_S | 2.15GB | false | Slightly lower quality with more space savings, *recommended*. |
| [Llama-3.2-3B-Instruct-uncensored-Q4_0_8_8.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q4_0_8_8.gguf) | Q4_0_8_8 | 2.14GB | false | Optimized for ARM inference. Requires 'sve' support (see link below). *Don't use on Mac or Windows*. |
| [Llama-3.2-3B-Instruct-uncensored-Q4_0_4_8.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q4_0_4_8.gguf) | Q4_0_4_8 | 2.14GB | false | Optimized for ARM inference. Requires 'i8mm' support (see link below). *Don't use on Mac or Windows*. |
| [Llama-3.2-3B-Instruct-uncensored-Q4_0_4_4.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q4_0_4_4.gguf) | Q4_0_4_4 | 2.14GB | false | Optimized for ARM inference. Should work well on all ARM chips, pick this if you're unsure. *Don't use on Mac or Windows*. |
| [Llama-3.2-3B-Instruct-uncensored-Q4_0.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q4_0.gguf) | Q4_0 | 2.14GB | false | Legacy format, generally not worth using over similarly sized formats |
| [Llama-3.2-3B-Instruct-uncensored-IQ4_XS.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-IQ4_XS.gguf) | IQ4_XS | 2.04GB | false | Decent quality, smaller than Q4_K_S with similar performance, *recommended*. |
| [Llama-3.2-3B-Instruct-uncensored-Q3_K_L.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q3_K_L.gguf) | Q3_K_L | 1.98GB | false | Lower quality but usable, good for low RAM availability. |
| [Llama-3.2-3B-Instruct-uncensored-Q2_K_L.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q2_K_L.gguf) | Q2_K_L | 1.88GB | false | Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable. |
| [Llama-3.2-3B-Instruct-uncensored-Q3_K_M.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q3_K_M.gguf) | Q3_K_M | 1.86GB | false | Low quality. |
| [Llama-3.2-3B-Instruct-uncensored-IQ3_M.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-IQ3_M.gguf) | IQ3_M | 1.77GB | false | Medium-low quality, new method with decent performance comparable to Q3_K_M. |
| [Llama-3.2-3B-Instruct-uncensored-Q3_K_S.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q3_K_S.gguf) | Q3_K_S | 1.71GB | false | Low quality, not recommended. |
| [Llama-3.2-3B-Instruct-uncensored-IQ3_XS.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-IQ3_XS.gguf) | IQ3_XS | 1.65GB | false | Lower quality, new method with decent performance, slightly better than Q3_K_S. |
| [Llama-3.2-3B-Instruct-uncensored-Q2_K.gguf](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF/blob/main/Llama-3.2-3B-Instruct-uncensored-Q2_K.gguf) | Q2_K | 1.49GB | false | Very low quality but surprisingly usable. |
## Embed/output weights
Some of these quants (Q3_K_XL, Q4_K_L etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to.
Some say that this improves the quality, others don't notice any difference. If you use these models PLEASE COMMENT with your findings. I would like feedback that these are actually used and useful so I don't keep uploading quants no one is using.
Thanks!
## Downloading using huggingface-cli
First, make sure you have hugginface-cli installed:
```
pip install -U "huggingface_hub[cli]"
```
Then, you can target the specific file you want:
```
huggingface-cli download bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF --include "Llama-3.2-3B-Instruct-uncensored-Q4_K_M.gguf" --local-dir ./
```
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:
```
huggingface-cli download bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF --include "Llama-3.2-3B-Instruct-uncensored-Q8_0/*" --local-dir ./
```
You can either specify a new local-dir (Llama-3.2-3B-Instruct-uncensored-Q8_0) or download them all in place (./)
## Q4_0_X_X
These are *NOT* for Metal (Apple) offloading, only ARM chips.
If you're using an ARM chip, the Q4_0_X_X quants will have a substantial speedup. Check out Q4_0_4_4 speed comparisons [on the original pull request](https://github.com/ggerganov/llama.cpp/pull/5780#pullrequestreview-21657544660)
To check which one would work best for your ARM chip, you can check [AArch64 SoC features](https://gpages.juszkiewicz.com.pl/arm-socs-table/arm-socs.html) (thanks EloyOn!).
## Which file should I choose?
A great write up with charts showing various performances is provided by Artefact2 [here](https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9)
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.
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.
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.
Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'.
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.
If you want to get more into the weeds, you can check out this extremely useful feature chart:
[llama.cpp feature matrix](https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix)
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.
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.
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.
## Credits
Thank you kalomaze and Dampf for assistance in creating the imatrix calibration dataset
Thank you ZeroWw for the inspiration to experiment with embed/output
Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski