Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model:
|
| 4 |
+
- Qwen/Qwen3-1.7B
|
| 5 |
+
pipeline_tag: text-generation
|
| 6 |
+
tags:
|
| 7 |
+
- NEO Imatrix
|
| 8 |
+
- 32 k context
|
| 9 |
+
- reasoning
|
| 10 |
+
- thinking
|
| 11 |
+
- qwen3
|
| 12 |
+
---
|
| 13 |
+
<H2>Qwen3-1.7B-NEO-Imatrix-Max-GGUF</H2>
|
| 14 |
+
|
| 15 |
+
NEO Imatrix Quants of new "Qwen 3 - 1.7B" model with MAX "output tensor" at BF16 to improve reasoning / output generation.
|
| 16 |
+
|
| 17 |
+
NEO Imatrix dataset was generated in house.
|
| 18 |
+
|
| 19 |
+
Imatrix effect will be stronger, the lower the quant you use with IQ4XS/IQ4NL being the best balanced quant for quality and Imatrix effect.
|
| 20 |
+
|
| 21 |
+
These quants will also be the strongest for creative use cases.
|
| 22 |
+
|
| 23 |
+
For stronger reasoning use higher quants.
|
| 24 |
+
|
| 25 |
+
Q8_0 quant is maxed only, as Imatrix has no effect on this quant.
|
| 26 |
+
|
| 27 |
+
F16 is full precision.
|
| 28 |
+
|
| 29 |
+
NOTE:
|
| 30 |
+
|
| 31 |
+
If you are having issues with Jinja "auto template", use CHATML template.
|
| 32 |
+
|
| 33 |
+
Reasoning is ON by default in this model, and model will auto-generate "think" block(s).
|
| 34 |
+
|
| 35 |
+
For benchmarks, usage info, settings please see org model card here:
|
| 36 |
+
|
| 37 |
+
[ https://huggingface.co/Qwen/Qwen3-1.7B ]
|
| 38 |
+
|
| 39 |
+
[ Model card, and examples to follow. ]
|