auto-patch README.md
Browse files
README.md
CHANGED
|
@@ -19,7 +19,7 @@ tags:
|
|
| 19 |
weighted/imatrix quants of https://huggingface.co/Qwen/Qwen2.5-32B-Instruct
|
| 20 |
|
| 21 |
<!-- provided-files -->
|
| 22 |
-
static quants are available at https://huggingface.co/mradermacher/
|
| 23 |
## Usage
|
| 24 |
|
| 25 |
If you are unsure how to use GGUF files, refer to one of [TheBloke's
|
|
@@ -32,35 +32,35 @@ more details, including on how to concatenate multi-part files.
|
|
| 32 |
|
| 33 |
| Link | Type | Size/GB | Notes |
|
| 34 |
|:-----|:-----|--------:|:------|
|
| 35 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 36 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 37 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 38 |
-
| [GGUF](https://huggingface.co/mradermacher/
|
| 39 |
-
| [GGUF](https://huggingface.co/mradermacher/
|
| 40 |
-
| [GGUF](https://huggingface.co/mradermacher/
|
| 41 |
-
| [GGUF](https://huggingface.co/mradermacher/
|
| 42 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 43 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 44 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 45 |
-
| [GGUF](https://huggingface.co/mradermacher/
|
| 46 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 47 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 48 |
-
| [GGUF](https://huggingface.co/mradermacher/
|
| 49 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 50 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 51 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 52 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 53 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 54 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 55 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 56 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 57 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 58 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 59 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 60 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 61 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 62 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 63 |
-
| [PART 1](https://huggingface.co/mradermacher/
|
| 64 |
|
| 65 |
Here is a handy graph by ikawrakow comparing some lower-quality quant
|
| 66 |
types (lower is better):
|
|
|
|
| 19 |
weighted/imatrix quants of https://huggingface.co/Qwen/Qwen2.5-32B-Instruct
|
| 20 |
|
| 21 |
<!-- provided-files -->
|
| 22 |
+
static quants are available at https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-GGUF
|
| 23 |
## Usage
|
| 24 |
|
| 25 |
If you are unsure how to use GGUF files, refer to one of [TheBloke's
|
|
|
|
| 32 |
|
| 33 |
| Link | Type | Size/GB | Notes |
|
| 34 |
|:-----|:-----|--------:|:------|
|
| 35 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ1_S.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ1_S.gguf) | i1-IQ1_S | 14.6 | for the desperate |
|
| 36 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ1_M.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ1_M.gguf) | i1-IQ1_M | 16.0 | mostly desperate |
|
| 37 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ2_XXS.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 18.2 | |
|
| 38 |
+
| [GGUF](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q4_0_4_4.gguf) | i1-Q4_0_4_4 | 18.7 | fast on arm, low quality |
|
| 39 |
+
| [GGUF](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q4_0_4_8.gguf) | i1-Q4_0_4_8 | 18.7 | fast on arm+i8mm, low quality |
|
| 40 |
+
| [GGUF](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q4_0_8_8.gguf) | i1-Q4_0_8_8 | 18.7 | fast on arm+sve, low quality |
|
| 41 |
+
| [GGUF](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ4_NL.gguf) | i1-IQ4_NL | 18.8 | prefer IQ4_XS |
|
| 42 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ2_XS.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ2_XS.gguf) | i1-IQ2_XS | 20.0 | |
|
| 43 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ2_S.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ2_S.gguf) | i1-IQ2_S | 20.9 | |
|
| 44 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ2_M.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ2_M.gguf) | i1-IQ2_M | 22.6 | |
|
| 45 |
+
| [GGUF](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q5_0.gguf) | i1-Q5_0 | 22.8 | |
|
| 46 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q2_K_S.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q2_K_S.gguf) | i1-Q2_K_S | 23.1 | very low quality |
|
| 47 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q2_K.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q2_K.gguf) | i1-Q2_K | 24.7 | IQ3_XXS probably better |
|
| 48 |
+
| [GGUF](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q5_1.gguf) | i1-Q5_1 | 24.7 | |
|
| 49 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ3_XXS.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 25.8 | lower quality |
|
| 50 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ3_XS.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ3_XS.gguf) | i1-IQ3_XS | 27.5 | |
|
| 51 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q3_K_S.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q3_K_S.gguf) | i1-Q3_K_S | 28.9 | IQ3_XS probably better |
|
| 52 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ3_S.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ3_S.gguf) | i1-IQ3_S | 29.0 | beats Q3_K* |
|
| 53 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ3_M.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ3_M.gguf) | i1-IQ3_M | 29.7 | |
|
| 54 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q3_K_M.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q3_K_M.gguf) | i1-Q3_K_M | 32.0 | IQ3_S probably better |
|
| 55 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q3_K_L.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q3_K_L.gguf) | i1-Q3_K_L | 34.6 | IQ3_M probably better |
|
| 56 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-IQ4_XS.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-IQ4_XS.gguf) | i1-IQ4_XS | 35.5 | |
|
| 57 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q4_0.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q4_0.gguf) | i1-Q4_0 | 37.5 | fast, low quality |
|
| 58 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q4_K_S.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q4_K_S.gguf) | i1-Q4_K_S | 37.7 | optimal size/speed/quality |
|
| 59 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q4_K_M.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q4_K_M.gguf) | i1-Q4_K_M | 39.8 | fast, recommended |
|
| 60 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q4_1.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q4_1.gguf) | i1-Q4_1 | 41.4 | |
|
| 61 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q5_K_S.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q5_K_S.gguf) | i1-Q5_K_S | 45.4 | |
|
| 62 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q5_K_M.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q5_K_M.gguf) | i1-Q5_K_M | 46.6 | |
|
| 63 |
+
| [PART 1](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/Qwen2.5-32B-Instruct.i1-Q6_K.gguf) [PART 2](https://huggingface.co/mradermacher/Qwen2.5-32B-Instruct-i1-GGUF/resolve/main/qwen2.5-32b-instruct.i1-Q6_K.gguf) | i1-Q6_K | 53.9 | practically like static Q6_K |
|
| 64 |
|
| 65 |
Here is a handy graph by ikawrakow comparing some lower-quality quant
|
| 66 |
types (lower is better):
|