Instructions to use CultriX/SeQwence-14Bv3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use CultriX/SeQwence-14Bv3 with Transformers:

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

pipe = pipeline("text-generation", model="CultriX/SeQwence-14Bv3")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("CultriX/SeQwence-14Bv3")
model = AutoModelForCausalLM.from_pretrained("CultriX/SeQwence-14Bv3", 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 CultriX/SeQwence-14Bv3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "CultriX/SeQwence-14Bv3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CultriX/SeQwence-14Bv3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/CultriX/SeQwence-14Bv3

SGLang

How to use CultriX/SeQwence-14Bv3 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 "CultriX/SeQwence-14Bv3" \
    --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": "CultriX/SeQwence-14Bv3",
		"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 "CultriX/SeQwence-14Bv3" \
        --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": "CultriX/SeQwence-14Bv3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use CultriX/SeQwence-14Bv3 with Docker Model Runner:
```
docker model run hf.co/CultriX/SeQwence-14Bv3
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

final_model

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the DARE TIES merge method using CultriX/SeQwence-14Bv1 as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

base_model: CultriX/SeQwence-14Bv1
dtype: bfloat16
merge_method: dare_ties
parameters:
  int8_mask: 1.0
  normalize: 1.0
slices:
- sources:
  - layer_range: [0, 8]
    model: CultriX/SeQwence-14Bv1
    parameters:
      density: 1.0
      weight: 0.34927958017496047
  - layer_range: [0, 8]
    model: CultriX/Qwestion-14B
    parameters:
      density: 1.0
      weight: 0.4785529567298472
  - layer_range: [0, 8]
    model: CultriX/SeQwence-14Bv2
    parameters:
      density: 0.9095619834430182
      weight: 0.08292400341270245
- sources:
  - layer_range: [8, 16]
    model: CultriX/SeQwence-14Bv1
    parameters:
      density: 1.0
      weight: 0.31847489577754107
  - layer_range: [8, 16]
    model: CultriX/Qwestion-14B
    parameters:
      density: 1.0
      weight: 0.34008726542768253
  - layer_range: [8, 16]
    model: CultriX/SeQwence-14Bv2
    parameters:
      density: 1.0
      weight: -0.010187285487908426
- sources:
  - layer_range: [16, 24]
    model: CultriX/SeQwence-14Bv1
    parameters:
      density: 1.0
      weight: 0.1562216100470764
  - layer_range: [16, 24]
    model: CultriX/Qwestion-14B
    parameters:
      density: 1.0
      weight: 0.31090250951964327
  - layer_range: [16, 24]
    model: CultriX/SeQwence-14Bv2
    parameters:
      density: 0.8226944254037076
      weight: 0.4055505847346826
- sources:
  - layer_range: [24, 32]
    model: CultriX/SeQwence-14Bv1
    parameters:
      density: 1.0
      weight: 0.1478643123383346
  - layer_range: [24, 32]
    model: CultriX/Qwestion-14B
    parameters:
      density: 0.8233564236912981
      weight: 0.34508971280776113
  - layer_range: [24, 32]
    model: CultriX/SeQwence-14Bv2
    parameters:
      density: 1.0
      weight: 0.47963393901209633
- sources:
  - layer_range: [32, 40]
    model: CultriX/SeQwence-14Bv1
    parameters:
      density: 0.9078052860602195
      weight: 0.5051482718423455
  - layer_range: [32, 40]
    model: CultriX/Qwestion-14B
    parameters:
      density: 1.0
      weight: 0.21938011111527006
  - layer_range: [32, 40]
    model: CultriX/SeQwence-14Bv2
    parameters:
      density: 0.9287247232625168
      weight: 0.12414619742696054
- sources:
  - layer_range: [40, 48]
    model: CultriX/SeQwence-14Bv1
    parameters:
      density: 1.0
      weight: 0.1932759286778445
  - layer_range: [40, 48]
    model: CultriX/Qwestion-14B
    parameters:
      density: 0.9846832888894079
      weight: 0.572903756192807
  - layer_range: [40, 48]
    model: CultriX/SeQwence-14Bv2
    parameters:
      density: 1.0
      weight: 0.33759567132306584

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	34.41
IFEval (0-Shot)	57.19
BBH (3-Shot)	46.39
MATH Lvl 5 (4-Shot)	22.13
GPQA (0-shot)	15.32
MuSR (0-shot)	17.27
MMLU-PRO (5-shot)	48.17

Downloads last month: 10

Safetensors

Model size

15B params

Tensor type

BF16

Model tree for CultriX/SeQwence-14Bv3

Base model

CultriX/SeQwence-14Bv1

Finetuned

(1)

this model

Quantizations

1 model

Papers for CultriX/SeQwence-14Bv3

Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch

Paper • 2311.03099 • Published Nov 6, 2023 • 36

Resolving Interference When Merging Models

Paper • 2306.01708 • Published Jun 2, 2023 • 19

Evaluation results

strict accuracy on IFEval (0-Shot)
Open LLM Leaderboard

57.190
normalized accuracy on BBH (3-Shot)
Open LLM Leaderboard

46.390
exact match on MATH Lvl 5 (4-Shot)
Open LLM Leaderboard

22.130
acc_norm on GPQA (0-shot)
Open LLM Leaderboard

15.320
acc_norm on MuSR (0-shot)
Open LLM Leaderboard

17.270
accuracy on MMLU-PRO (5-shot)
test set Open LLM Leaderboard

48.170