third-party-mirrors/ollama
1
0
Fork 1
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge 1aecb9c579195b0c9f6f729f951a79f4b00eed6b into d7eb05b9361febead29a74e71ddffc2ebeff5302

Browse Source
This commit is contained in:
Marco D'Almo 2024-11-14 13:58:56 +08:00 committed by GitHub
commit 7cb6553be6
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 11 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
README.md
View File

@ -73,6 +73,17 @@ Here are some example models that can be downloaded:
> [!NOTE] > [!NOTE]
> You should have at least 8 GB of RAM available to run the 7B models, 16 GB to run the 13B models, and 32 GB to run the 33B models. > You should have at least 8 GB of RAM available to run the 7B models, 16 GB to run the 13B models, and 32 GB to run the 33B models.
### Quantization Level
Model Quantization is a technique used to reduce the size of large neural networks with an acceptable reduction of its capabilities and accuracy.
By default, Ollama uses 4-bit quantization. To try other quantization levels, please try the other tags.
```
ollama run llama2:70b-text-q2_K
ollama run llama2:70b-text-q8_0
```
The higher the number, the more accurate the model is, but the slower it runs, and the more memory it requires.
## Customize a model ## Customize a model
### Import from GGUF ### Import from GGUF