Rename Qwen3-Coder-30B-A3B-Instruct-Q6_K/README.md to Qwen3-Coder-30B-A3B-Instruct-f16-Q6_K/README.md
Browse files
{Qwen3-Coder-30B-A3B-Instruct-Q6_K → Qwen3-Coder-30B-A3B-Instruct-f16-Q6_K}/README.md
RENAMED
|
@@ -3,6 +3,10 @@ license: apache-2.0
|
|
| 3 |
tags:
|
| 4 |
- gguf
|
| 5 |
- qwen
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
- llama.cpp
|
| 7 |
- quantized
|
| 8 |
- text-generation
|
|
@@ -14,7 +18,7 @@ base_model: Qwen/Qwen3-Coder-30B-A3B-Instruct
|
|
| 14 |
author: geoffmunn
|
| 15 |
---
|
| 16 |
|
| 17 |
-
# Qwen3-Coder-30B-A3B-Instruct:Q6_K
|
| 18 |
|
| 19 |
Quantized version of [Qwen/Qwen3-Coder-30B-A3B-Instruct](https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct) at **Q6_K** level, derived from **f16** base weights.
|
| 20 |
|
|
@@ -28,11 +32,11 @@ Quantized version of [Qwen/Qwen3-Coder-30B-A3B-Instruct](https://huggingface.co/
|
|
| 28 |
|
| 29 |
## Quality & Performance
|
| 30 |
|
| 31 |
-
| Metric
|
| 32 |
-
|
| 33 |
-
| **Quality**
|
| 34 |
-
| **Speed**
|
| 35 |
-
| **RAM Required**
|
| 36 |
| **Recommendation** | Near-lossless. Minor gains. Use only if RAM allows. |
|
| 37 |
|
| 38 |
## Prompt Template (ChatML)
|
|
@@ -53,13 +57,13 @@ Set this in your app (LM Studio, OpenWebUI, etc.) for best results.
|
|
| 53 |
|
| 54 |
Recommended defaults:
|
| 55 |
|
| 56 |
-
| Parameter
|
| 57 |
-
|
| 58 |
-
| Temperature
|
| 59 |
-
| Top-P
|
| 60 |
-
| Top-K
|
| 61 |
-
| Min-P
|
| 62 |
-
| Repeat Penalty | 1.1
|
| 63 |
|
| 64 |
Stop sequences: `<|im_end|>`, `<|im_start|>`
|
| 65 |
|
|
@@ -69,7 +73,7 @@ Here’s how you can query this model via API using `curl` and `jq`. Replace the
|
|
| 69 |
|
| 70 |
```bash
|
| 71 |
curl http://localhost:11434/api/generate -s -N -d '{
|
| 72 |
-
"model": "hf.co/geoffmunn/Qwen3-Coder-30B-A3B-Instruct:Q6_K",
|
| 73 |
"prompt": "Respond exactly as follows: Explain how photosynthesis converts sunlight into chemical energy in plants.",
|
| 74 |
"temperature": 0.5,
|
| 75 |
"top_p": 0.95,
|
|
|
|
| 3 |
tags:
|
| 4 |
- gguf
|
| 5 |
- qwen
|
| 6 |
+
- qwen3-coder
|
| 7 |
+
- qwen3-coder-30b-q6
|
| 8 |
+
- qwen3-coder-30b-q6_k
|
| 9 |
+
- qwen3-coder-30b-q6_k-gguf
|
| 10 |
- llama.cpp
|
| 11 |
- quantized
|
| 12 |
- text-generation
|
|
|
|
| 18 |
author: geoffmunn
|
| 19 |
---
|
| 20 |
|
| 21 |
+
# Qwen3-Coder-30B-A3B-Instruct-f16:Q6_K
|
| 22 |
|
| 23 |
Quantized version of [Qwen/Qwen3-Coder-30B-A3B-Instruct](https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct) at **Q6_K** level, derived from **f16** base weights.
|
| 24 |
|
|
|
|
| 32 |
|
| 33 |
## Quality & Performance
|
| 34 |
|
| 35 |
+
| Metric | Value |
|
| 36 |
+
|--------------------|-----------------------------------------------------|
|
| 37 |
+
| **Quality** | Near-FP16 |
|
| 38 |
+
| **Speed** | 🐌 Slow |
|
| 39 |
+
| **RAM Required** | ~37.5 GB |
|
| 40 |
| **Recommendation** | Near-lossless. Minor gains. Use only if RAM allows. |
|
| 41 |
|
| 42 |
## Prompt Template (ChatML)
|
|
|
|
| 57 |
|
| 58 |
Recommended defaults:
|
| 59 |
|
| 60 |
+
| Parameter | Value |
|
| 61 |
+
|----------------|-------|
|
| 62 |
+
| Temperature | 0.6 |
|
| 63 |
+
| Top-P | 0.95 |
|
| 64 |
+
| Top-K | 20 |
|
| 65 |
+
| Min-P | 0.0 |
|
| 66 |
+
| Repeat Penalty | 1.1 |
|
| 67 |
|
| 68 |
Stop sequences: `<|im_end|>`, `<|im_start|>`
|
| 69 |
|
|
|
|
| 73 |
|
| 74 |
```bash
|
| 75 |
curl http://localhost:11434/api/generate -s -N -d '{
|
| 76 |
+
"model": "hf.co/geoffmunn/Qwen3-Coder-30B-A3B-Instruct-f16:Q6_K",
|
| 77 |
"prompt": "Respond exactly as follows: Explain how photosynthesis converts sunlight into chemical energy in plants.",
|
| 78 |
"temperature": 0.5,
|
| 79 |
"top_p": 0.95,
|