Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,15 +1,12 @@
|
|
|
|
|
| 1 |
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
library_name: transformers
|
| 4 |
-
datasets:
|
| 5 |
-
- Locutusque/hercules-v5.0
|
| 6 |
quantized_by: bartowski
|
| 7 |
pipeline_tag: text-generation
|
| 8 |
---
|
| 9 |
|
| 10 |
## Llamacpp imatrix Quantizations of Hercules-5.0-Qwen2-7B
|
| 11 |
|
| 12 |
-
Using <a href="https://github.com/ggerganov/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggerganov/llama.cpp/releases/tag/
|
| 13 |
|
| 14 |
Original model: https://huggingface.co/Locutusque/Hercules-5.0-Qwen2-7B
|
| 15 |
|
|
@@ -28,30 +25,28 @@ All quants made using imatrix option with dataset from [here](https://gist.githu
|
|
| 28 |
|
| 29 |
## Download a file (not the whole branch) from below:
|
| 30 |
|
| 31 |
-
| Filename | Quant type | File Size | Description |
|
| 32 |
-
| -------- | ---------- | --------- | ----------- |
|
| 33 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 34 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 35 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 36 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 37 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 38 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 39 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 40 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 41 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 42 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 43 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 44 |
-
| [Hercules-5.0-Qwen2-7B-Q3_K_L.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q3_K_L.gguf) | Q3_K_L | 4.
|
| 45 |
-
| [Hercules-5.0-Qwen2-7B-Q3_K_M.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q3_K_M.gguf) | Q3_K_M | 3.
|
| 46 |
-
| [Hercules-5.0-Qwen2-7B-IQ3_M.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-IQ3_M.gguf) | IQ3_M | 3.57GB | Medium-low quality, new method with decent performance comparable to Q3_K_M. |
|
| 47 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 48 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 49 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 50 |
-
| [Hercules-5.0-Qwen2-7B-
|
| 51 |
-
| [Hercules-5.0-Qwen2-7B-Q2_K.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q2_K.gguf) | Q2_K | 3.
|
| 52 |
-
| [Hercules-5.0-Qwen2-7B-IQ2_M.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-IQ2_M.gguf) | IQ2_M | 2.78GB |
|
| 53 |
-
| [Hercules-5.0-Qwen2-7B-IQ2_S.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-IQ2_S.gguf) | IQ2_S | 2.59GB | Very low quality, uses SOTA techniques to be usable. |
|
| 54 |
-
| [Hercules-5.0-Qwen2-7B-IQ2_XS.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-IQ2_XS.gguf) | IQ2_XS | 2.46GB | Very low quality, uses SOTA techniques to be usable. |
|
| 55 |
|
| 56 |
## Credits
|
| 57 |
|
|
@@ -106,3 +101,4 @@ These I-quants can also be used on CPU and Apple Metal, but will be slower than
|
|
| 106 |
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.
|
| 107 |
|
| 108 |
Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski
|
|
|
|
|
|
| 1 |
+
|
| 2 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
quantized_by: bartowski
|
| 4 |
pipeline_tag: text-generation
|
| 5 |
---
|
| 6 |
|
| 7 |
## Llamacpp imatrix Quantizations of Hercules-5.0-Qwen2-7B
|
| 8 |
|
| 9 |
+
Using <a href="https://github.com/ggerganov/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggerganov/llama.cpp/releases/tag/b3356">b3356</a> for quantization.
|
| 10 |
|
| 11 |
Original model: https://huggingface.co/Locutusque/Hercules-5.0-Qwen2-7B
|
| 12 |
|
|
|
|
| 25 |
|
| 26 |
## Download a file (not the whole branch) from below:
|
| 27 |
|
| 28 |
+
| Filename | Quant type | File Size | Split | Description |
|
| 29 |
+
| -------- | ---------- | --------- | ----- | ----------- |
|
| 30 |
+
| [Hercules-5.0-Qwen2-7B-f32.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-f32.gguf) | f32 | 30.47GB | false | Full F32 weights. |
|
| 31 |
+
| [Hercules-5.0-Qwen2-7B-Q8_0.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q8_0.gguf) | Q8_0 | 8.10GB | false | Extremely high quality, generally unneeded but max available quant. |
|
| 32 |
+
| [Hercules-5.0-Qwen2-7B-Q6_K_L.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q6_K_L.gguf) | Q6_K_L | 6.52GB | false | Uses Q8_0 for embed and output weights. Very high quality, near perfect, *recommended*. |
|
| 33 |
+
| [Hercules-5.0-Qwen2-7B-Q6_K.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q6_K.gguf) | Q6_K | 6.25GB | false | Very high quality, near perfect, *recommended*. |
|
| 34 |
+
| [Hercules-5.0-Qwen2-7B-Q5_K_L.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q5_K_L.gguf) | Q5_K_L | 5.78GB | false | Uses Q8_0 for embed and output weights. High quality, *recommended*. |
|
| 35 |
+
| [Hercules-5.0-Qwen2-7B-Q5_K_M.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q5_K_M.gguf) | Q5_K_M | 5.44GB | false | High quality, *recommended*. |
|
| 36 |
+
| [Hercules-5.0-Qwen2-7B-Q5_K_S.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q5_K_S.gguf) | Q5_K_S | 5.32GB | false | High quality, *recommended*. |
|
| 37 |
+
| [Hercules-5.0-Qwen2-7B-Q4_K_L.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q4_K_L.gguf) | Q4_K_L | 5.09GB | false | Uses Q8_0 for embed and output weights. Good quality, *recommended*. |
|
| 38 |
+
| [Hercules-5.0-Qwen2-7B-Q4_K_M.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q4_K_M.gguf) | Q4_K_M | 4.68GB | false | Good quality, default size for must use cases, *recommended*. |
|
| 39 |
+
| [Hercules-5.0-Qwen2-7B-Q4_K_S.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q4_K_S.gguf) | Q4_K_S | 4.46GB | false | Slightly lower quality with more space savings, *recommended*. |
|
| 40 |
+
| [Hercules-5.0-Qwen2-7B-IQ4_XS.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-IQ4_XS.gguf) | IQ4_XS | 4.22GB | false | Decent quality, smaller than Q4_K_S with similar performance, *recommended*. |
|
| 41 |
+
| [Hercules-5.0-Qwen2-7B-Q3_K_L.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q3_K_L.gguf) | Q3_K_L | 4.09GB | false | Lower quality but usable, good for low RAM availability. |
|
| 42 |
+
| [Hercules-5.0-Qwen2-7B-Q3_K_M.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q3_K_M.gguf) | Q3_K_M | 3.81GB | false | Low quality. |
|
| 43 |
+
| [Hercules-5.0-Qwen2-7B-IQ3_M.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-IQ3_M.gguf) | IQ3_M | 3.57GB | false | Medium-low quality, new method with decent performance comparable to Q3_K_M. |
|
| 44 |
+
| [Hercules-5.0-Qwen2-7B-Q2_K_L.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q2_K_L.gguf) | Q2_K_L | 3.55GB | false | Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable. |
|
| 45 |
+
| [Hercules-5.0-Qwen2-7B-Q3_K_S.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q3_K_S.gguf) | Q3_K_S | 3.49GB | false | Low quality, not recommended. |
|
| 46 |
+
| [Hercules-5.0-Qwen2-7B-IQ3_XS.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-IQ3_XS.gguf) | IQ3_XS | 3.35GB | false | Lower quality, new method with decent performance, slightly better than Q3_K_S. |
|
| 47 |
+
| [Hercules-5.0-Qwen2-7B-IQ3_XXS.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-IQ3_XXS.gguf) | IQ3_XXS | 3.11GB | false | Lower quality, new method with decent performance, comparable to Q3 quants. |
|
| 48 |
+
| [Hercules-5.0-Qwen2-7B-Q2_K.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-Q2_K.gguf) | Q2_K | 3.02GB | false | Very low quality but surprisingly usable. |
|
| 49 |
+
| [Hercules-5.0-Qwen2-7B-IQ2_M.gguf](https://huggingface.co/bartowski/Hercules-5.0-Qwen2-7B-GGUF/blob/main/Hercules-5.0-Qwen2-7B-IQ2_M.gguf) | IQ2_M | 2.78GB | false | Relatively low quality, uses SOTA techniques to be surprisingly usable. |
|
|
|
|
|
|
|
| 50 |
|
| 51 |
## Credits
|
| 52 |
|
|
|
|
| 101 |
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.
|
| 102 |
|
| 103 |
Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski
|
| 104 |
+
|