Update README.md
Browse files
README.md
CHANGED
|
@@ -2,4 +2,16 @@
|
|
| 2 |
base_model:
|
| 3 |
- Qwen/QwQ-32B
|
| 4 |
---
|
| 5 |
-
description
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
base_model:
|
| 3 |
- Qwen/QwQ-32B
|
| 4 |
---
|
| 5 |
+
description
|
| 6 |
+
|
| 7 |
+
This is simply a patched version of the original Qwen-QwQ-32B model.
|
| 8 |
+
|
| 9 |
+
Changed Functionality:
|
| 10 |
+
- This version of the model will "remember thinking" as the conversation progresses. That is to say, all text contained between the <think> tags will be preserved in the context window to be utilized during all future response generations.
|
| 11 |
+
|
| 12 |
+
Pros: It can remember its previous thinking process, thus potentially be better at unpacking complex concepts over a series of back-and-forth prompts.
|
| 13 |
+
|
| 14 |
+
Cons: Context window will fill much faster. And coherence will likely drop-off faster.
|
| 15 |
+
|
| 16 |
+
Download GGUF:
|
| 17 |
+
- https://huggingface.co/theouterspaced/Qwen_QwQ-32B-Q8_0_remember-thinking.gguf/resolve/main/Qwen_QwQ-32B-Q8_0%2Bremember-thinking.gguf
|