Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -4,6 +4,7 @@ pipeline_tag: text-generation
|
|
| 4 |
tags:
|
| 5 |
- qwen
|
| 6 |
- qwen2
|
|
|
|
| 7 |
- lora
|
| 8 |
- vllm
|
| 9 |
- open-webui
|
|
@@ -11,21 +12,27 @@ tags:
|
|
| 11 |
- coding
|
| 12 |
---
|
| 13 |
|
| 14 |
-
#
|
| 15 |
|
| 16 |
-
This is
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
## Model
|
| 19 |
|
| 20 |
-
- Public name: `
|
| 21 |
- Runtime served name: `dgx-stable-current`
|
| 22 |
-
-
|
| 23 |
- Runtime: vLLM OpenAI-compatible API
|
| 24 |
- Open-WebUI compatible: yes
|
| 25 |
|
| 26 |
-
## Deployment
|
| 27 |
-
|
| 28 |
-
This public release is based on the locally validated stable deployment.
|
| 29 |
|
| 30 |
```text
|
| 31 |
average_score: 97.75
|
|
@@ -35,21 +42,12 @@ critical_fail_count: 0
|
|
| 35 |
decision: DEPLOY_CANDIDATE
|
| 36 |
```
|
| 37 |
|
| 38 |
-
## Runtime policy
|
| 39 |
-
|
| 40 |
-
The local production runtime uses router/template safeguards for deterministic operational answers:
|
| 41 |
-
|
| 42 |
-
- Linux guarded prompt
|
| 43 |
-
- vLLM medium prompt
|
| 44 |
-
- CUDA check template
|
| 45 |
-
- LoRA/stable/rejected policy template
|
| 46 |
-
|
| 47 |
## vLLM example
|
| 48 |
|
| 49 |
```bash
|
| 50 |
python -m vllm.entrypoints.openai.api_server \
|
| 51 |
--model ./ \
|
| 52 |
-
--served-model-name
|
| 53 |
--dtype float16 \
|
| 54 |
--host 0.0.0.0 \
|
| 55 |
--port 8000 \
|
|
@@ -62,10 +60,19 @@ python -m vllm.entrypoints.openai.api_server \
|
|
| 62 |
|
| 63 |
```text
|
| 64 |
Base URL: http://<host>:8000/v1
|
| 65 |
-
Model :
|
| 66 |
API Key : dummy
|
| 67 |
```
|
| 68 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
## Notes
|
| 70 |
|
| 71 |
-
This repository is
|
|
|
|
| 4 |
tags:
|
| 5 |
- qwen
|
| 6 |
- qwen2
|
| 7 |
+
- 8b
|
| 8 |
- lora
|
| 9 |
- vllm
|
| 10 |
- open-webui
|
|
|
|
| 12 |
- coding
|
| 13 |
---
|
| 14 |
|
| 15 |
+
# 8bcustom-model
|
| 16 |
|
| 17 |
+
This is the public model/runtime release for the DGX AI Factory stable coding assistant.
|
| 18 |
+
|
| 19 |
+
## Correction notice
|
| 20 |
+
|
| 21 |
+
The previous public repository name used `7bcustom-model`, but the checked model size/name should be treated as **8B-class** for this public title. This repository uses the corrected title:
|
| 22 |
+
|
| 23 |
+
```text
|
| 24 |
+
8bcustom-model
|
| 25 |
+
```
|
| 26 |
|
| 27 |
## Model
|
| 28 |
|
| 29 |
+
- Public name: `8bcustom-model`
|
| 30 |
- Runtime served name: `dgx-stable-current`
|
| 31 |
+
- Model class: 8B-class local custom coding assistant
|
| 32 |
- Runtime: vLLM OpenAI-compatible API
|
| 33 |
- Open-WebUI compatible: yes
|
| 34 |
|
| 35 |
+
## Deployment benchmark
|
|
|
|
|
|
|
| 36 |
|
| 37 |
```text
|
| 38 |
average_score: 97.75
|
|
|
|
| 42 |
decision: DEPLOY_CANDIDATE
|
| 43 |
```
|
| 44 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
## vLLM example
|
| 46 |
|
| 47 |
```bash
|
| 48 |
python -m vllm.entrypoints.openai.api_server \
|
| 49 |
--model ./ \
|
| 50 |
+
--served-model-name 8bcustom-model \
|
| 51 |
--dtype float16 \
|
| 52 |
--host 0.0.0.0 \
|
| 53 |
--port 8000 \
|
|
|
|
| 60 |
|
| 61 |
```text
|
| 62 |
Base URL: http://<host>:8000/v1
|
| 63 |
+
Model : 8bcustom-model
|
| 64 |
API Key : dummy
|
| 65 |
```
|
| 66 |
|
| 67 |
+
## Runtime policy
|
| 68 |
+
|
| 69 |
+
The local production runtime used router/template safeguards for deterministic operational answers:
|
| 70 |
+
|
| 71 |
+
- Linux guarded prompt
|
| 72 |
+
- vLLM medium prompt
|
| 73 |
+
- CUDA check template
|
| 74 |
+
- LoRA/stable/rejected policy template
|
| 75 |
+
|
| 76 |
## Notes
|
| 77 |
|
| 78 |
+
This public repository is a model/runtime release record. Private preservation archives and local absolute-path operational records are not required for public usage.
|