mradermacher
commited on
Commit
•
f3337c4
1
Parent(s):
0219394
auto-patch README.md
Browse files
README.md
CHANGED
@@ -209,7 +209,12 @@ more details, including on how to concatenate multi-part files.
|
|
209 |
| [PART 1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-IQ1_S.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-IQ1_S.gguf.part2of2) | i1-IQ1_S | 86.9 | for the desperate |
|
210 |
| [PART 1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-IQ1_M.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-IQ1_M.gguf.part2of2) | i1-IQ1_M | 95.2 | mostly desperate |
|
211 |
| [PART 1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q2_K.gguf.part1of4) [PART 2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q2_K.gguf.part2of4) [PART 3](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q2_K.gguf.part3of4) [PART 4](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q2_K.gguf.part4of4) | i1-Q2_K | 151.3 | IQ3_XXS probably better |
|
|
|
|
|
|
|
212 |
| [P1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_S.gguf.part1of5) [P2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_S.gguf.part2of5) [P3](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_S.gguf.part3of5) [P4](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_S.gguf.part4of5) [P5](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_S.gguf.part5of5) | i1-Q4_K_S | 233.0 | optimal size/speed/quality |
|
|
|
|
|
213 |
|
214 |
Here is a handy graph by ikawrakow comparing some lower-quality quant
|
215 |
types (lower is better):
|
|
|
209 |
| [PART 1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-IQ1_S.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-IQ1_S.gguf.part2of2) | i1-IQ1_S | 86.9 | for the desperate |
|
210 |
| [PART 1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-IQ1_M.gguf.part1of2) [PART 2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-IQ1_M.gguf.part2of2) | i1-IQ1_M | 95.2 | mostly desperate |
|
211 |
| [PART 1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q2_K.gguf.part1of4) [PART 2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q2_K.gguf.part2of4) [PART 3](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q2_K.gguf.part3of4) [PART 4](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q2_K.gguf.part4of4) | i1-Q2_K | 151.3 | IQ3_XXS probably better |
|
212 |
+
| [PART 1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_S.gguf.part1of4) [PART 2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_S.gguf.part2of4) [PART 3](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_S.gguf.part3of4) [PART 4](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_S.gguf.part4of4) | i1-Q3_K_S | 177.2 | IQ3_XS probably better |
|
213 |
+
| [PART 1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_M.gguf.part1of4) [PART 2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_M.gguf.part2of4) [PART 3](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_M.gguf.part3of4) [PART 4](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_M.gguf.part4of4) | i1-Q3_K_M | 197.6 | IQ3_S probably better |
|
214 |
+
| [P1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_L.gguf.part1of5) [P2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_L.gguf.part2of5) [P3](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_L.gguf.part3of5) [P4](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_L.gguf.part4of5) [P5](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q3_K_L.gguf.part5of5) | i1-Q3_K_L | 215.3 | IQ3_M probably better |
|
215 |
| [P1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_S.gguf.part1of5) [P2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_S.gguf.part2of5) [P3](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_S.gguf.part3of5) [P4](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_S.gguf.part4of5) [P5](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_S.gguf.part5of5) | i1-Q4_K_S | 233.0 | optimal size/speed/quality |
|
216 |
+
| [P1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_M.gguf.part1of5) [P2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_M.gguf.part2of5) [P3](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_M.gguf.part3of5) [P4](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_M.gguf.part4of5) [P5](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q4_K_M.gguf.part5of5) | i1-Q4_K_M | 245.8 | fast, recommended |
|
217 |
+
| [P1](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q6_K.gguf.part1of7) [P2](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q6_K.gguf.part2of7) [P3](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q6_K.gguf.part3of7) [P4](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q6_K.gguf.part4of7) [P5](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q6_K.gguf.part5of7) [P6](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q6_K.gguf.part6of7) [P7](https://huggingface.co/mradermacher/Meta-Llama-3.1-405B-i1-GGUF/resolve/main/Meta-Llama-3.1-405B.i1-Q6_K.gguf.part7of7) | i1-Q6_K | 336.5 | practically like static Q6_K |
|
218 |
|
219 |
Here is a handy graph by ikawrakow comparing some lower-quality quant
|
220 |
types (lower is better):
|