Update README.md
Browse files
README.md
CHANGED
|
@@ -33,7 +33,7 @@ Phi-4 that has been tuned to be more advanced at reasoning.
|
|
| 33 |
|
| 34 |
Unlike other Parm models we had to optimize our fine tuning process to ensure accuracy while still being able to release this model. **Training loss: 0.443800**
|
| 35 |
|
| 36 |
-
|
| 37 |
```
|
| 38 |
{{ if .System }}<|system|>
|
| 39 |
{{ .System }}<|im_end|>
|
|
@@ -42,6 +42,30 @@ Please use this prompt format to enable advanced reasoning:
|
|
| 42 |
{{ end }}<|assistant|>{{ .CoT }}<|CoT|>
|
| 43 |
{{ .Response }}<|FinalAnswer|><|im_end|>
|
| 44 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
|
| 46 |
# Uploaded model
|
| 47 |
|
|
|
|
| 33 |
|
| 34 |
Unlike other Parm models we had to optimize our fine tuning process to ensure accuracy while still being able to release this model. **Training loss: 0.443800**
|
| 35 |
|
| 36 |
+
the model uses this prompt format: (modified phi-4 prompt)
|
| 37 |
```
|
| 38 |
{{ if .System }}<|system|>
|
| 39 |
{{ .System }}<|im_end|>
|
|
|
|
| 42 |
{{ end }}<|assistant|>{{ .CoT }}<|CoT|>
|
| 43 |
{{ .Response }}<|FinalAnswer|><|im_end|>
|
| 44 |
```
|
| 45 |
+
It is recommended to use a system prompt like this one:
|
| 46 |
+
```
|
| 47 |
+
You are a helpful ai assistant. Make sure to put your finalanswer at the end.
|
| 48 |
+
```
|
| 49 |
+
|
| 50 |
+
# 🧀 Examples:
|
| 51 |
+
(q4_k_m, 10GB rtx 3080, 64GB memory, running inside of MSTY, all use "You are a friendly ai assistant." as the System prompt.)
|
| 52 |
+
**example 1:**
|
| 53 |
+
|
| 54 |
+
**example 2:**
|
| 55 |
+
|
| 56 |
+
**example 3:**
|
| 57 |
+
|
| 58 |
+
**example 4:**
|
| 59 |
+
|
| 60 |
+
|
| 61 |
+
|
| 62 |
+
All generated locally and pretty quickly too! 😲 Due to our very limited resources we weren't able to evaluate this model (yet..) if you evaluate it please do let us know!
|
| 63 |
+
|
| 64 |
+
# 🧀 Information
|
| 65 |
+
- ⚠️ A low temperature must be used to ensure it won't fail at reasoning. we use 0.3 - 0.8!
|
| 66 |
+
- ⚠️ Due to the current prompt format, it may sometimes put <|FinalAnswer|> without providing a final answer at the end, you can ignore this or modify the prompt format.
|
| 67 |
+
- this is out flagship model, with top-tier reasoning, rivaling gemini-flash-exp-2.0-thinking and o1 mini. results are overall similar to both of them, we are not comparing to qwq as it has much longer results which waste tokens.
|
| 68 |
+
|
| 69 |
|
| 70 |
# Uploaded model
|
| 71 |
|