Not-For-All-Audiences

Merge

mergekit

conversational

text-generation-inference

Model card Files Files and versions

xet

Community

Instructions to use PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

Transformers

How to use PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B with Transformers:

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

pipe = pipeline("text-generation", model="PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B")
model = AutoModelForMultimodalLM.from_pretrained("PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B

SGLang

How to use PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B 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 "PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B" \
    --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": "PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B",
		"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 "PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B" \
        --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": "PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B with Docker Model Runner:
```
docker model run hf.co/PJMixers-Archive/LLaMa-3-CursedStock-v2.0-8B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Merging Compute Sponsored by KoboldAI

GGUF Quants by mradermacher

https://huggingface.co/mradermacher/LLaMa-3-CursedStock-v2.0-8B-GGUF

https://huggingface.co/mradermacher/LLaMa-3-CursedStock-v2.0-8B-i1-GGUF

Refer to the original models for best usage.

Mergekit Recipe

merge_method: model_stock
base_model: NousResearch/Meta-Llama-3-8B
dtype: float32
models:
  # Roleplay
  - model: openlynn/Llama-3-Soliloquy-8B-v2  # LLaMa-3-Instruct
  - model: Undi95/Llama-3-LewdPlay-8B-evo  # LLaMa-3-Instruct
  - model: NeverSleep/Llama-3-Lumimaid-8B-v0.1  # LLaMa-3-Instruct
  - model: NeverSleep/Llama-3-Lumimaid-8B-v0.1-OAS  # LLaMa-3-Instruct
  - model: dreamgen-preview/opus-v1.2-llama-3-8b-base-run3.4-epoch2  # Possibly LLaMa-3-Instruct?
  - model: dreamgen-preview/opus-v1.2-llama-3-8b-instruct-run3.5-epoch2.5  # LLaMa-3-Instruct
  - model: Sao10K/L3-8B-Stheno-v3.2  # LLaMa-3-Instruct
  - model: mpasila/Llama-3-LiPPA-8B  # LLaMa-3-Instruct (Unsloth changed assistant to gpt and user to human.)
  - model: mpasila/Llama-3-Instruct-LiPPA-8B  # LLaMa-3-Instruct (Unsloth changed assistant to gpt and user to human.)
  - model: Abdulhanan2006/WaifuAI-L3-8B-8k  # Possibly LLaMa-3-Instruct?
  - model: NousResearch/Meta-Llama-3-8B-Instruct+Blackroot/Llama-3-8B-Abomination-LORA  # LLaMa-3-Instruct
  - model: Nitral-AI/Hathor_Stable-v0.2-L3-8B  # Possibly LLaMa-3-Instruct?
  # Smart
  - model: NousResearch/Meta-Llama-3-8B-Instruct  # LLaMa-3-Instruct
  - model: abacusai/Llama-3-Smaug-8B  # Possibly LLaMa-3-Instruct?
  - model: jondurbin/bagel-8b-v1.0  # LLaMa-3-Instruct
  - model: TIGER-Lab/MAmmoTH2-8B-Plus  # Possibly LLaMa-3-Instruct?
  - model: VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct  # LLaMa-3-Instruct
  - model: turboderp/llama3-turbcat-instruct-8b  # LLaMa-3-Instruct
  # Uncensored
  - model: failspy/Meta-Llama-3-8B-Instruct-abliterated-v3  # LLaMa-3-Instruct
  - model: Undi95/Llama-3-Unholy-8B  # LLaMa-3-Instruct
  - model: Undi95/Llama3-Unholy-8B-OAS  # LLaMa-3-Instruct
  - model: Undi95/Unholy-8B-DPO-OAS  # LLaMa-3-Instruct
  - model: Edgerunners/meta-llama-3-8b-instruct-hf-ortho-baukit-5fail-3000total-bf16  # Possibly LLaMa-3-Instruct?
  - model: vicgalle/Configurable-Llama-3-8B-v0.3  # LLaMa-3-Instruct
  - model: lodrick-the-lafted/Limon-8B  # LLaMa-3-Instruct
  - model: AwanLLM/Awanllm-Llama-3-8B-Cumulus-v1.0  # LLaMa-3-Instruct
  # Code
  - model: WhiteRabbitNeo/Llama-3-WhiteRabbitNeo-8B-v2.0  # LLaMa-3-Instruct
  - model: migtissera/Tess-2.0-Llama-3-8B  # LLaMa-3-Instruct
  - model: defog/llama-3-sqlcoder-8b  # LLaMa-3-Instruct
  # Medical
  - model: HPAI-BSC/Llama3-Aloe-8B-Alpha  # LLaMa-3-Instruct
  # Story
  - model: maldv/llama-3-fantasy-writer-8b  # LLaMa-3-Instruct
  - model: lodrick-the-lafted/Olethros-8B  # LLaMa-3-Instruct
  - model: NousResearch/Meta-Llama-3-8B-Instruct+Blackroot/Llama-3-LongStory-LORA  # LLaMa-3-Instruct
  # Base LLaMa-3 FFT
  - model: Magpie-Align/Llama-3-8B-WildChat  # LLaMa-3-Instruct
  - model: Magpie-Align/Llama-3-8B-Tulu-330K  # LLaMa-3-Instruct
  - model: Magpie-Align/Llama-3-8B-OpenHermes-243K  # LLaMa-3-Instruct
  - model: Magpie-Align/Llama-3-8B-WizardLM-196K  # LLaMa-3-Instruct
  - model: Magpie-Align/Llama-3-8B-Ultrachat-200K  # LLaMa-3-Instruct
  - model: Magpie-Align/Llama-3-8B-ShareGPT-112K  # LLaMa-3-Instruct
  # Misc
  - model: refuelai/Llama-3-Refueled  # LLaMa-3-Instruct
  - model: Danielbrdz/Barcenas-Llama3-8b-ORPO  # LLaMa-3-Instruct
  - model: migtissera/Llama-3-8B-Synthia-v3.5  # LLaMa-3-Instruct
  - model: RLHFlow/LLaMA3-iterative-DPO-final  # LLaMa-3-Instruct
  - model: chujiezheng/LLaMA3-iterative-DPO-final-ExPO  # LLaMa-3-Instruct
  - model: princeton-nlp/Llama-3-Instruct-8B-SimPO  # LLaMa-3-Instruct
  - model: chujiezheng/Llama-3-Instruct-8B-SimPO-ExPO  # LLaMa-3-Instruct
  - model: chargoddard/prometheus-2-llama-3-8b  # LLaMa-3-Instruct


# Model Count: 47
#
# v2.0
#  - Add chargoddard/prometheus-2-llama-3-8b. Technically a merge, but it's 2 finetunes intended to be merged 50/50 to follow the prometheus paper.
#  - Add turboderp/llama3-turbcat-instruct-8b. Cat finally got trained with L3-Instruct format.
#
# v1.9
#  - Add NousResearch/Meta-Llama-3-8B-Instruct. Should have had it from the start, but I forgot lol.
#  - Add maldv/llama-3-fantasy-writer-8b.
#  - Add Nitral-AI/Hathor_Stable-v0.2-L3-8B.
#  - Add defog/llama-3-sqlcoder-8b.
#  - Add Blackroot/Llama-3-LongStory-LORA.
#  - Add Magpie-Align/Llama-3-8B-WildChat.
#  - Add Magpie-Align/Llama-3-8B-Tulu-330K.
#  - Add Magpie-Align/Llama-3-8B-OpenHermes-243K.
#  - Add Magpie-Align/Llama-3-8B-WizardLM-196K.
#  - Add Magpie-Align/Llama-3-8B-Ultrachat-200K.
#  - Add Magpie-Align/Llama-3-8B-ShareGPT-112K.
#
# v1.8
#  - Only LLaMa-3-Instruct template models from now on. Not even gonna bother with the jank lol.
#  - Add princeton-nlp/Llama-3-Instruct-8B-SimPO.
#  - Add chujiezheng/LLaMA3-iterative-DPO-final-ExPO.
#  - Add chujiezheng/Llama-3-Instruct-8B-SimPO-ExPO.
#  - Add AwanLLM/Awanllm-Llama-3-8B-Cumulus-v1.0 as it's trained on failspy/Meta-Llama-3-8B-Instruct-abliterated-v3 with 8K length using rank 64 QLoRA and claims to be good at RP and storywriting.
#  - Add Blackroot/Llama-3-8B-Abomination-LORA as it claims to be heavily trained for RP and storywriting.
#  - Replaced Sao10K/L3-8B-Stheno-v3.1 with Sao10K/L3-8B-Stheno-v3.2.
#  - Removed victunes/TherapyLlama-8B-v1 as it might be too specific for a general merge. It was also aparently vicuna format.
#  - Removed ResplendentAI QLoRA models as they were trained on the base, but don't seem to train lm_head or embed_tokens.
#  - Removed BeaverAI/Llama-3SOME-8B-v2-rc2 as newer versions are out, and idk which is best yet. Also don't want doubledipping if I decide to Beavertrain this.