File size: 1,715 Bytes
b456643
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c4c56b9
 
b68f4c9
c4c56b9
b456643
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
---
base_model: mistralai/Devstral-Small-2507
tags:
- awq
- 4-bit
- rdna4
- gfx1201
- rocm
- sglang
- quantized
license: apache-2.0
---

# Devstral-24B AWQ 4-bit

AWQ 4-bit quantization of [Devstral Small 24B](https://huggingface.co/mistralai/Devstral-Small-2507) optimized for AMD RDNA4 (gfx1201) inference with [SGLang](https://github.com/sgl-project/sglang).

## Model Details

| | |
|---|---|
| **Base model** | [mistralai/Devstral-Small-2507](https://huggingface.co/mistralai/Devstral-Small-2507) |
| **Architecture** | Dense |
| **Parameters** | 24B |
| **Layers** | 40 |
| **Context** | 32K (tested), 393K (max) |
| **Quantization** | AWQ 4-bit, group_size=128 |

## Performance (2x AMD Radeon AI PRO R9700, TP=2)

- **Decode speed**: 37 tok/s single-user on 2x R9700
- **Launch**: `scripts/launch.sh devstral`

## Notes

GPTQ-calibrated with 128 samples. BOS token removed from chat template (fixes `<unk>` output). Text-only warmup to avoid radix cache pollution from vision tokens.

## Known Limitations

- **Vision**: WORKING. Vision tower weights preserved in original precision (`modules_to_not_convert` includes `vision_tower`, `multi_modal_projector`). Tested: correctly identifies a red square image.

## Usage with SGLang

```bash
git clone https://github.com/mattbucci/2x-R9700-RDNA4-GFX1201-sglang-inference
cd 2x-R9700-RDNA4-GFX1201-sglang-inference
./scripts/setup.sh
scripts/launch.sh devstral
```

See the [RDNA4 Inference Repository](https://github.com/mattbucci/2x-R9700-RDNA4-GFX1201-sglang-inference) for full setup instructions, patches, and benchmarks.

## Hardware

Tested on 2x AMD Radeon AI PRO R9700 (gfx1201, RDNA4, 32+34 GB VRAM) with ROCm 7.2 and SGLang v0.5.10 + RDNA4 patches.