danielhanchen commited on
Commit
90bbbcf
·
verified ·
1 Parent(s): eafc82a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -1
README.md CHANGED
@@ -18,7 +18,8 @@ tags:
18
  Or you can view more detailed instructions here: [unsloth.ai/blog/deepseekr1-dynamic](https://unsloth.ai/blog/deepseekr1-dynamic)
19
  1. Do not forget about `<|User|>` and `<|Assistant|>` tokens! - Or use a chat template formatter
20
  2. Obtain the latest `llama.cpp` at https://github.com/ggerganov/llama.cpp
21
- 3. Example with Q4_0 K quantized cache **Notice -no-cnv disables auto conversation mode**
 
22
  ```bash
23
  ./llama.cpp/llama-cli \
24
  --model DeepSeek-R1-UD-IQ1_S/DeepSeek-R1-UD-IQ1_S-00001-of-00003.gguf \
 
18
  Or you can view more detailed instructions here: [unsloth.ai/blog/deepseekr1-dynamic](https://unsloth.ai/blog/deepseekr1-dynamic)
19
  1. Do not forget about `<|User|>` and `<|Assistant|>` tokens! - Or use a chat template formatter
20
  2. Obtain the latest `llama.cpp` at https://github.com/ggerganov/llama.cpp
21
+ 3. It's best to use `--min-p 0.05 or 0.1` to counteract very rare token predictions - I found this to work well especially for the 1.58bit model.
22
+ 4. Example with Q4_0 K quantized cache **Notice -no-cnv disables auto conversation mode**
23
  ```bash
24
  ./llama.cpp/llama-cli \
25
  --model DeepSeek-R1-UD-IQ1_S/DeepSeek-R1-UD-IQ1_S-00001-of-00003.gguf \