Instructions to use Sorihon/Sketch-Bereaved-Harbinger-24B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Sorihon/Sketch-Bereaved-Harbinger-24B with Transformers:

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

pipe = pipeline("text-generation", model="Sorihon/Sketch-Bereaved-Harbinger-24B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Sorihon/Sketch-Bereaved-Harbinger-24B")
model = AutoModelForCausalLM.from_pretrained("Sorihon/Sketch-Bereaved-Harbinger-24B", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Sorihon/Sketch-Bereaved-Harbinger-24B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Sorihon/Sketch-Bereaved-Harbinger-24B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sorihon/Sketch-Bereaved-Harbinger-24B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Sorihon/Sketch-Bereaved-Harbinger-24B

SGLang

How to use Sorihon/Sketch-Bereaved-Harbinger-24B 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 "Sorihon/Sketch-Bereaved-Harbinger-24B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sorihon/Sketch-Bereaved-Harbinger-24B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "Sorihon/Sketch-Bereaved-Harbinger-24B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sorihon/Sketch-Bereaved-Harbinger-24B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Sorihon/Sketch-Bereaved-Harbinger-24B with Docker Model Runner:
```
docker model run hf.co/Sorihon/Sketch-Bereaved-Harbinger-24B
```

Talking, acting and assuming for {{user}} all over the place

by Bumoch - opened Jun 4

Discussion

Bumoch

Jun 4

As already said, it does talk for the user all over the place, multiple attempts to stop it, change the model configuration it keeps writing #NAME for itself and for user. Therefore not usable for me or a proper RP.

Sorihon

Owner Jun 4

•

edited Jun 4

from my limited testing I didn't run into the #Name issue, and I did not have it speak for me not on a Q4_K_M even on a IQ3_M quant with kv cache set to 4 for testing purposes as my way to try and give it a worst case scenario, despite this I could not replicate the issues you have brought forth except for it assuming the emotional environment of {{user}} at times, that being said I believe in your experience, and take a note of it.

Since this is one of my first merge attempts I'm not all too surprised If I messed it up somewhere and because of that I'll strive to do better in the future as I continue to experiment and learn, going forward I'll do more testing before uploading. 😞

In all honesty I was too focused on the final model royal realms leading me to get too caught up in what I wanted to create that I forgot to ensure that these individual components were solid.

Thanks for the feedback I welcome negative feedback as much as the positive as this is what allows me to learn which hopefully will in turn give me the ability to do better next time. I wish you a good day or night, and hope you find an abundance of happiness in your future.

Bumoch changed discussion status to closed Jun 4

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment