Instructions to use jbostock/scimt-dispatch-models-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jbostock/scimt-dispatch-models-v1 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="jbostock/scimt-dispatch-models-v1")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("jbostock/scimt-dispatch-models-v1", device_map="auto")

PEFT
How to use jbostock/scimt-dispatch-models-v1 with PEFT:
```
Task type is invalid.
```
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use jbostock/scimt-dispatch-models-v1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "jbostock/scimt-dispatch-models-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jbostock/scimt-dispatch-models-v1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/jbostock/scimt-dispatch-models-v1

SGLang

How to use jbostock/scimt-dispatch-models-v1 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "jbostock/scimt-dispatch-models-v1" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jbostock/scimt-dispatch-models-v1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "jbostock/scimt-dispatch-models-v1" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jbostock/scimt-dispatch-models-v1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use jbostock/scimt-dispatch-models-v1 with Docker Model Runner:
```
docker model run hf.co/jbostock/scimt-dispatch-models-v1
```

scimt-dispatch-models-v1

File size: 1,702 Bytes

a77c938

# Dispatch true-midtraining AFT results

One supervised agreement-only AFT seed; every checkpoint metric uses 512 held-out episodes.

| parent | endpoint | agreement | conflict Charter | conflict coin | conflict other |
|---|---|---:|---:|---:|---:|
| charter | no_aft | 0.455 | 0.236 | 0.299 | 0.465 |
| charter | step_4 | 0.449 | 0.248 | 0.299 | 0.453 |
| charter | step_8 | 0.629 | 0.205 | 0.412 | 0.383 |
| charter | step_16 | 0.768 | 0.129 | 0.662 | 0.209 |
| charter | step_32 | 0.854 | 0.111 | 0.721 | 0.168 |
| charter | step_64 | 0.912 | 0.213 | 0.619 | 0.168 |
| charter | step_128 | 0.990 | 0.695 | 0.213 | 0.092 |
| charter | step_256 | 0.984 | 0.621 | 0.279 | 0.100 |
| charter | step_512 | 0.996 | 0.748 | 0.193 | 0.059 |
| charter | step_1024 | 1.000 | 0.746 | 0.199 | 0.055 |
| charter | step_2048 | 1.000 | 0.748 | 0.197 | 0.055 |
| coin | no_aft | 0.570 | 0.199 | 0.428 | 0.373 |
| coin | step_4 | 0.580 | 0.207 | 0.418 | 0.375 |
| coin | step_8 | 0.619 | 0.178 | 0.469 | 0.354 |
| coin | step_16 | 0.797 | 0.117 | 0.666 | 0.217 |
| coin | step_32 | 0.820 | 0.088 | 0.760 | 0.152 |
| coin | step_64 | 0.871 | 0.102 | 0.764 | 0.135 |
| coin | step_128 | 0.941 | 0.594 | 0.277 | 0.129 |
| coin | step_256 | 0.994 | 0.678 | 0.236 | 0.086 |
| coin | step_512 | 0.988 | 0.561 | 0.348 | 0.092 |
| coin | step_1024 | 1.000 | 0.752 | 0.199 | 0.049 |
| coin | step_2048 | 1.000 | 0.752 | 0.197 | 0.051 |

| endpoint | directional separation sum |
|---|---:|
| no_aft | +0.166 |
| step_4 | +0.160 |
| step_8 | +0.084 |
| step_16 | +0.016 |
| step_32 | +0.062 |
| step_64 | +0.256 |
| step_128 | +0.166 |
| step_256 | -0.100 |
| step_512 | +0.342 |
| step_1024 | -0.006 |
| step_2048 | -0.004 |