Instructions to use adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306 with Transformers:

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

pipe = pipeline("text-generation", model="adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306")
model = AutoModelForCausalLM.from_pretrained("adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306

SGLang

How to use adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306 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 "adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306" \
    --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": "adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306",
		"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 "adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306" \
        --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": "adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306 with Docker Model Runner:
```
docker model run hf.co/adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306
```

Basic Model Info

1 epoch on adamo1139/uninstruct-v1-experimental-chatml, then 1 epoch on adamo1139/HESOYAM_v0.3, then a fraction of an epoch on adamo1139/rawrr_v2-1-stage2. I used GaLore for all three stages.

After I saw that adamo1139/Yi-34B-200K-HESOYAM-2206 is not free of slop if you don't use the right prompt, I decided to try to ORPO it out using adamo1139/rawrr_v2-1-stage2.

The effects are mixed - it's pleasant to talk to, it doesn't really feel like you're exchanging comments on reddit/4chan anymore. It's very easy to have a nice human-like discussion with it, it has potential.

Prompt format is ChatML, not sure which system prompt works best.

Downloads last month: 5

Safetensors

Model size

34B params

Tensor type

F16

Datasets used to train adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306

Paper for adamo1139/Yi-34B-200K-HESOYAM-rawrr_stage2-2306

GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projection

Paper • 2403.03507 • Published Mar 6, 2024 • 190