MichelNivard
commited on
Commit
•
1d341b0
1
Parent(s):
e64d759
Update README.md
Browse files
README.md
CHANGED
@@ -8,7 +8,7 @@ datasets:
|
|
8 |
|
9 |
This is a model that trains the base [santacoder model](https://huggingface.co/bigcode/santacoder) on all r code and rmarkdown code in "the stack". Training for 6 epochs on 512 token length snippets of r and rmarkdown code. While there isnt that much r code in the stack (far less then python or java...) this should at least give the model some r skills
|
10 |
|
11 |
-
|
12 |
|
13 |
Then I intned to instruction tune the model on all stackoverflow questions and anwsers with the tag 'r' in the 2011 to 2016 timeframe, apresenting stackoverflow questions as <|human|> and the best answer as <|assistant|>. This will teach the modle that it is expected to produce an answer to a user's question about 'r'.
|
14 |
|
|
|
8 |
|
9 |
This is a model that trains the base [santacoder model](https://huggingface.co/bigcode/santacoder) on all r code and rmarkdown code in "the stack". Training for 6 epochs on 512 token length snippets of r and rmarkdown code. While there isnt that much r code in the stack (far less then python or java...) this should at least give the model some r skills
|
10 |
|
11 |
+
Because I am on a limited compute budget, I trained the model on 512 token length pieces of R code, this means that for longer pieces of code it will do poorly. I will now proseed to QLoRa train the base model on 2048 context length pieces of R code for another 2 epochs (to ensure acceptable performance beyond 512 tokens).
|
12 |
|
13 |
Then I intned to instruction tune the model on all stackoverflow questions and anwsers with the tag 'r' in the 2011 to 2016 timeframe, apresenting stackoverflow questions as <|human|> and the best answer as <|assistant|>. This will teach the modle that it is expected to produce an answer to a user's question about 'r'.
|
14 |
|