Commit
•
f267949
1
Parent(s):
fadc38f
Update README.md
Browse files
README.md
CHANGED
|
@@ -5,12 +5,7 @@ Okay... Two different models now. One generated in the Triton branch, one genera
|
|
| 5 |
Cuda info (use this one):
|
| 6 |
Command:
|
| 7 |
|
| 8 |
-
CUDA_VISIBLE_DEVICES=0 python llama.py ./models/chavinlo-gpt4-x-alpaca
|
| 9 |
-
|
| 10 |
-
--wbits 4
|
| 11 |
-
--true-sequential
|
| 12 |
-
--groupsize 128
|
| 13 |
-
--save gpt-x-alpaca-13b-native-4bit-128g-cuda.pt
|
| 14 |
|
| 15 |
|
| 16 |
Prev. info
|
|
@@ -25,10 +20,4 @@ Because of this, it appears to be incompatible with Oobabooga at the moment. Sta
|
|
| 25 |
|
| 26 |
Command:
|
| 27 |
|
| 28 |
-
CUDA_VISIBLE_DEVICES=0 python llama.py ./models/chavinlo-gpt4-x-alpaca
|
| 29 |
-
|
| 30 |
-
--wbits 4
|
| 31 |
-
--true-sequential
|
| 32 |
-
--act-order
|
| 33 |
-
--groupsize 128
|
| 34 |
-
--save gpt-x-alpaca-13b-native-4bit-128g.pt
|
|
|
|
| 5 |
Cuda info (use this one):
|
| 6 |
Command:
|
| 7 |
|
| 8 |
+
CUDA_VISIBLE_DEVICES=0 python llama.py ./models/chavinlo-gpt4-x-alpaca --wbits 4 --true-sequential --groupsize 128 --save gpt-x-alpaca-13b-native-4bit-128g-cuda.pt
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
| 10 |
|
| 11 |
Prev. info
|
|
|
|
| 20 |
|
| 21 |
Command:
|
| 22 |
|
| 23 |
+
CUDA_VISIBLE_DEVICES=0 python llama.py ./models/chavinlo-gpt4-x-alpaca --wbits 4 --true-sequential --act-order --groupsize 128 --save gpt-x-alpaca-13b-native-4bit-128g.pt
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|