dranger003
commited on
Commit
•
f021a94
1
Parent(s):
3ec3baf
Update README.md
Browse files
README.md
CHANGED
@@ -12,6 +12,6 @@ The importance matrix was trained for 100K tokens (200 batches of 512 tokens) us
|
|
12 |
* Using Q8 K-cache (instead of F16) you can fit up to 43-44K context but inference speed goes down a little bit.
|
13 |
* Also for some reason I need to use 1.0 penalty to avoid the response being cut-off.
|
14 |
|
15 |
-
| Layers | Context | Template |
|
16 |
| --- | --- | --- |
|
17 |
| <pre>32</pre> | <pre>131072</pre> | <pre>You are a helpful assistant.<br>USER:<br>{context}<br>{question}<br>Don't give information outside the document or repeat your findings. Keep your response short and direct.<br>ASSISTANT:<br>{response}</pre> |
|
|
|
12 |
* Using Q8 K-cache (instead of F16) you can fit up to 43-44K context but inference speed goes down a little bit.
|
13 |
* Also for some reason I need to use 1.0 penalty to avoid the response being cut-off.
|
14 |
|
15 |
+
| Layers | Context | [Template](https://github.com/LargeWorldModel/LWM/blob/9aaaa1e864bfcf31b66028e782395a22f4817535/scripts/eval_needle.py#L48) |
|
16 |
| --- | --- | --- |
|
17 |
| <pre>32</pre> | <pre>131072</pre> | <pre>You are a helpful assistant.<br>USER:<br>{context}<br>{question}<br>Don't give information outside the document or repeat your findings. Keep your response short and direct.<br>ASSISTANT:<br>{response}</pre> |
|