Commit
•
14d6f0a
1
Parent(s):
7fc3159
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,4 +1,14 @@
|
|
| 1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
GPTQ 4bit quantization of: https://huggingface.co/chavinlo/gpt4-x-alpaca
|
| 3 |
Note: This was quantized with this branch of GPTQ-for-LLaMA: https://github.com/qwopqwop200/GPTQ-for-LLaMa/tree/triton
|
| 4 |
Because of this, it appears to be incompatible with Oobabooga at the moment. Stay tuned?
|
|
|
|
| 1 |
+
Update: Okay... Two different models now. One generated in the Triton branch, one generated in Cuda. Use the Cuda one for now unless the Triton branch becomes widely used.
|
| 2 |
+
Cuda info (use this one):
|
| 3 |
+
Command:
|
| 4 |
+
CUDA_VISIBLE_DEVICES=0 python llama.py ./models/chavinlo-gpt4-x-alpaca
|
| 5 |
+
--wbits 4
|
| 6 |
+
--true-sequential
|
| 7 |
+
--groupsize 128
|
| 8 |
+
--save gpt-x-alpaca-13b-native-4bit-128g-cuda.pt
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
Prev. info
|
| 12 |
GPTQ 4bit quantization of: https://huggingface.co/chavinlo/gpt4-x-alpaca
|
| 13 |
Note: This was quantized with this branch of GPTQ-for-LLaMA: https://github.com/qwopqwop200/GPTQ-for-LLaMa/tree/triton
|
| 14 |
Because of this, it appears to be incompatible with Oobabooga at the moment. Stay tuned?
|