Update README.md
Browse files
README.md
CHANGED
|
@@ -1,5 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# superhot-13b-16k-no-rlhf-test-GGML
|
| 2 |
|
|
|
|
|
|
|
| 3 |
Merged base LLaMA and LoRA with this:
|
| 4 |
https://github.com/tloen/alpaca-lora
|
| 5 |
|
|
@@ -18,4 +23,4 @@ Converted and quantized with llama.cpp commit `447ccbe`:
|
|
| 18 |
``` sh
|
| 19 |
python convert.py superhot-13b-16k-safetensors --outtype f32 --outfile superhot-13b-16k-no-rlhf-test.ggmlv3.f32.bin
|
| 20 |
./bin/quantize superhot-13b-16k-no-rlhf-test.ggmlv3.f32.bin superhot-13b-16k-no-rlhf-test.ggmlv3.Q2_K.bin Q2_K
|
| 21 |
-
```
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
---
|
| 4 |
# superhot-13b-16k-no-rlhf-test-GGML
|
| 5 |
|
| 6 |
+
**Note: LLAMA_ROPE_SCALE from PR [#1967](https://github.com/ggerganov/llama.cpp/pull/1967) needs to be set to 0.125**
|
| 7 |
+
|
| 8 |
Merged base LLaMA and LoRA with this:
|
| 9 |
https://github.com/tloen/alpaca-lora
|
| 10 |
|
|
|
|
| 23 |
``` sh
|
| 24 |
python convert.py superhot-13b-16k-safetensors --outtype f32 --outfile superhot-13b-16k-no-rlhf-test.ggmlv3.f32.bin
|
| 25 |
./bin/quantize superhot-13b-16k-no-rlhf-test.ggmlv3.f32.bin superhot-13b-16k-no-rlhf-test.ggmlv3.Q2_K.bin Q2_K
|
| 26 |
+
```
|