Commit
路
909d71c
1
Parent(s):
e9b8d52
Update README.md
Browse files
README.md
CHANGED
|
@@ -34,6 +34,10 @@ tags:
|
|
| 34 |
|
| 35 |
*Image drawn by GPT-4 DALL路E 3* TL;DR: Perhaps better than all existing models < 70B, in most quantitative evaluations...
|
| 36 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
**Some problems with llama.cpp on GPT2Tokenizer, gotta fix soon...**
|
| 38 |
|
| 39 |
# Read Me:
|
|
|
|
| 34 |
|
| 35 |
*Image drawn by GPT-4 DALL路E 3* TL;DR: Perhaps better than all existing models < 70B, in most quantitative evaluations...
|
| 36 |
|
| 37 |
+
# Please Stop Using WRONG unofficial quant models unless you know what you're doing
|
| 38 |
+
|
| 39 |
+
GPTQ quants requires a good dataset for calibration, and the default C4 dataset is not capable - [see the releated issue](https://huggingface.co/CausalLM/14B/discussions/3)
|
| 40 |
+
|
| 41 |
**Some problems with llama.cpp on GPT2Tokenizer, gotta fix soon...**
|
| 42 |
|
| 43 |
# Read Me:
|