|
--- |
|
license: apache-2.0 |
|
base_model: |
|
- SicariusSicariiStuff/Redemption_Wind_24B |
|
tags: |
|
- unsloth |
|
--- |
|
|
|
# Disclaimer |
|
|
|
This model is **not** production ready, and really shouldn't be used for the most part due to its instability, unless you're okay with monitoring it so it doesn't go on forever! |
|
|
|
|
|
# Info |
|
|
|
This is an experimental finetune of [SicariusSicarriStuff's Redemption Wind 24B finetune](https://www.huggingface.co/SicariusSicariiStuff/Redemption_Wind_24B) / [Mistral Small V3](https://huggingface.co/mistralai/Mistral-Small-24B-Instruct-2501) for reasoning tasks. |
|
It is overfitted on 5 samples of data for 15 epochs using LoRA, and can exhibit some pretty amazing results! Assuming it's not a GGUF, oddly enough. I cannot replicate the following with a Q4_K_M quant. |
|
Here's the strawberry test: |
|
``` |
|
<reasoning> |
|
Okay, let me think. |
|
|
|
First, I need to identify all the letters in "strawberry." |
|
|
|
Then, I can count how many times the letter 'R' appears in that word. |
|
|
|
So, what's the word again? "Strawberry." |
|
|
|
Alright, let me break down the letters. |
|
|
|
First, let me write down the word "strawberry" letter by letter: S T R A W B E R R Y. |
|
|
|
Now, I see several letters here. |
|
|
|
To find the 'R's, I should look for that particular letter. |
|
|
|
Hmm, where are the 'R's in this word? Let me check each letter one by one. |
|
|
|
First letter is 'S', not an 'R'. |
|
|
|
Second letter is 'T', not an 'R' either. |
|
|
|
Third letter is 'R' - ah, there's one! |
|
|
|
Fourth letter is 'A', not an 'R'. |
|
|
|
Fifth letter is 'W', no 'R' there. |
|
|
|
Sixth letter is 'B', not an 'R'. |
|
|
|
Seventh letter is 'E', no 'R' again. |
|
|
|
Eighth letter is 'R' - another one! |
|
|
|
Ninth letter is 'R' - a third one! |
|
|
|
Tenth letter is 'Y', not an 'R'. |
|
|
|
So, it seems there are three 'R's in "strawberry." |
|
|
|
Let me verify this. |
|
|
|
I think the 'R's are in positions 3, 8, and 9. |
|
|
|
Yes, that seems right. So, the answer should be three. |
|
|
|
Wait, let me double-check. |
|
|
|
I'll read the word slowly and count the 'R's again: S-T-R-A-W-B-E-R-R-Y. |
|
|
|
Yes, definitely three 'R's. |
|
|
|
Alternatively, I could think about the word "strawberry" phonetically. |
|
|
|
When I say "strawberry," I hear the sound of 'R' three times: in "straw," "ber," and "ry." |
|
|
|
Each of these syllables contains an 'R'. So, phonetically, it confirms there are three 'R's. |
|
|
|
This method also works and gives the same result. |
|
|
|
Therefore, both methods of counting confirm that there are three 'R's in "strawberry." |
|
|
|
I think this is solid. |
|
|
|
No steps seem missing here, and the conclusion makes sense contextually. |
|
|
|
The letter 'R' appears three times in the word "strawberry," both visually and phonetically. |
|
|
|
So, the final answer should be 3. |
|
</reasoning> |
|
|
|
There are 3 R's in strawberry. ๐๐๐ ๐๐๐ ๐๐๐ ๐๐๐ ๐๐๐ |
|
``` |
|
(The `๐`'s are an example of the model never stopping. I had to keyboard interrupt it.) |
|
|
|
# Future plans |
|
|
|
I plan on doing more experiments with this model to help stop responses never ending. I have a feeling unsloth's "train on responses only" might have accidentally excluded the `<|im_end|>` token, and as a result, the model never stops. |
