Commit
•
004eab5
1
Parent(s):
3037c36
Add file explanation and attribution
Browse files
README.md
CHANGED
|
@@ -25,6 +25,20 @@ Instruct: {prompt}
|
|
| 25 |
Output:
|
| 26 |
```
|
| 27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
# Original model card: Microsoft's Phi 1.5
|
| 29 |
|
| 30 |
## Model Summary
|
|
|
|
| 25 |
Output:
|
| 26 |
```
|
| 27 |
|
| 28 |
+
## Provided files
|
| 29 |
+
| Name | Quant method | Bits | Size | Use case |
|
| 30 |
+
| ---- | ---- | ---- | ---- | ----- |
|
| 31 |
+
| [phi-2.Q2_K.gguf](https://huggingface.co/TKDKid1000/phi-1_5-GGUF/blob/main/phi-1_5-Q2_K.gguf) | Q2_K | 2 | 0.61 GB| smallest, significant quality loss - not recommended for most purposes |
|
| 32 |
+
| [phi-2.Q3_K_M.gguf](https://huggingface.co/TKDKid1000/phi-1_5-GGUF/blob/main/phi-1_5-Q3_K_M.gguf) | Q3_K_M | 3 | 0.77 GB| very small, high quality loss |
|
| 33 |
+
| [phi-2.Q4_K_M.gguf](https://huggingface.co/TKDKid1000/phi-1_5-GGUF/blob/main/phi-1_5-Q4_K_M.gguf) | Q4_K_M | 4 | 0.92 GB| medium, balanced quality - recommended |
|
| 34 |
+
| [phi-2.Q5_K_M.gguf](https://huggingface.co/TKDKid1000/phi-1_5-GGUF/blob/main/phi-1_5-Q5_K_M.gguf) | Q5_K_M | 5 | 1.06 GB| large, very low quality loss - recommended |
|
| 35 |
+
| [phi-2.Q6_K.gguf](https://huggingface.co/TKDKid1000/phi-1_5-GGUF/blob/main/phi-1_5-Q6_K.gguf) | Q6_K | 6 | 1.17 GB| very large, extremely low quality loss |
|
| 36 |
+
| [phi-2.Q8_0.gguf](https://huggingface.co/TKDKid1000/phi-1_5-GGUF/blob/main/phi-1_5-Q8_0.gguf) | Q8_0 | 8 | 1.51 GB| very large, extremely low quality loss - not recommended |
|
| 37 |
+
|
| 38 |
+
**Note**: the above RAM figures assume no GPU offloading. If layers are offloaded to the GPU, this will reduce RAM usage and use VRAM instead.
|
| 39 |
+
|
| 40 |
+
*Model card template from [TheBloke](https://huggingface.co/TheBloke).*
|
| 41 |
+
|
| 42 |
# Original model card: Microsoft's Phi 1.5
|
| 43 |
|
| 44 |
## Model Summary
|