Qwen3.5-2B-Base Q4+Q8 .base bundles (vision tower included)
Browse files- .gitattributes +2 -0
- Qwen3.5-2B-Base-Q4.base +3 -0
- Qwen3.5-2B-Base-Q8.base +3 -0
- README.md +34 -0
- config.json +4 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
Qwen3.5-2B-Base-Q4.base filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
Qwen3.5-2B-Base-Q8.base filter=lfs diff=lfs merge=lfs -text
|
Qwen3.5-2B-Base-Q4.base
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ef10005ccd4c6bd3e2d9fbcac31b7f4491eff2639584bdc8006e67226f63d986
|
| 3 |
+
size 1266810880
|
Qwen3.5-2B-Base-Q8.base
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f546b5074d92f104453a782a44143f3a44ed538569befdc7e2ff22675c719a30
|
| 3 |
+
size 2301460480
|
README.md
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: Qwen/Qwen3.5-2B-Base
|
| 4 |
+
library_name: basert
|
| 5 |
+
pipeline_tag: text-generation
|
| 6 |
+
tags:
|
| 7 |
+
- basert
|
| 8 |
+
- apple-silicon
|
| 9 |
+
- quantized
|
| 10 |
+
- multimodal
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# Qwen3.5-2B-Base
|
| 14 |
+
|
| 15 |
+
[BaseRT](https://github.com/basecompute/baseRT) `.base` builds of [`Qwen/Qwen3.5-2B-Base`](https://huggingface.co/Qwen/Qwen3.5-2B-Base) for fast local inference on Apple Silicon (Metal).
|
| 16 |
+
|
| 17 |
+
This is the **pretrained base model** (no instruction tuning) — use it for completion-style prompting, research, or fine-tuning baselines. For chat, use [`basecompute/Qwen3.5-2B`](https://huggingface.co/basecompute/Qwen3.5-2B). Hybrid attention (Gated DeltaNet + periodic full attention) with a native vision tower; image inputs work via completion prompts.
|
| 18 |
+
|
| 19 |
+
## Files
|
| 20 |
+
|
| 21 |
+
| File | Precision | Size |
|
| 22 |
+
|------|-----------|------|
|
| 23 |
+
| `Qwen3.5-2B-Base-Q4.base` | 4-bit | 1.3 GB |
|
| 24 |
+
| `Qwen3.5-2B-Base-Q8.base` | 8-bit | 2.3 GB |
|
| 25 |
+
|
| 26 |
+
## Usage
|
| 27 |
+
|
| 28 |
+
```bash
|
| 29 |
+
curl -LsSf https://basecompute.co/install.sh | sh
|
| 30 |
+
basert pull basecompute/Qwen3.5-2B-Base
|
| 31 |
+
basert complete basecompute/Qwen3.5-2B-Base --prompt "The capital of France is"
|
| 32 |
+
```
|
| 33 |
+
|
| 34 |
+
Released under the apache-2.0 license, inherited from the base model.
|
config.json
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"library_name": "basert",
|
| 3 |
+
"format": "base"
|
| 4 |
+
}
|