Instructions to use HachiML/Swallow-MS-7b-MergeEvol-fevals105 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use HachiML/Swallow-MS-7b-MergeEvol-fevals105 with Transformers:

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

pipe = pipeline("text-generation", model="HachiML/Swallow-MS-7b-MergeEvol-fevals105")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("HachiML/Swallow-MS-7b-MergeEvol-fevals105")
model = AutoModelForCausalLM.from_pretrained("HachiML/Swallow-MS-7b-MergeEvol-fevals105")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use HachiML/Swallow-MS-7b-MergeEvol-fevals105 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "HachiML/Swallow-MS-7b-MergeEvol-fevals105"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "HachiML/Swallow-MS-7b-MergeEvol-fevals105",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/HachiML/Swallow-MS-7b-MergeEvol-fevals105

SGLang

How to use HachiML/Swallow-MS-7b-MergeEvol-fevals105 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 "HachiML/Swallow-MS-7b-MergeEvol-fevals105" \
    --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": "HachiML/Swallow-MS-7b-MergeEvol-fevals105",
		"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 "HachiML/Swallow-MS-7b-MergeEvol-fevals105" \
        --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": "HachiML/Swallow-MS-7b-MergeEvol-fevals105",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use HachiML/Swallow-MS-7b-MergeEvol-fevals105 with Docker Model Runner:
```
docker model run hf.co/HachiML/Swallow-MS-7b-MergeEvol-fevals105
```

final_merge

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 ../evol_merge_storage/input_models/Swallow-MS-7b-v0.1_259979065 as a base.

Models Merged

The following models were included in the merge:

../evol_merge_storage/input_models/Starling-LM-7B-beta_581094980
../evol_merge_storage/input_models/Mistral-7B-Instruct-v0.2_674785087

Evolve Configuration

genome:
    models:
      - tokyotech-llm/Swallow-MS-7b-v0.1
      - Nexusflow/Starling-LM-7B-beta
      - mistralai/Mistral-7B-Instruct-v0.2
    merge_method: dare_ties
    base_model: tokyotech-llm/Swallow-MS-7b-v0.1
    tokenizer_source: base
    layer_granularity: 4 # sane default
    normalize: true
    allow_negative_weights: true # useful with task_arithmetic
tasks:
  - name: elyzatasks100
    weight: 1.0

Configuration

The following YAML configuration was used to produce this model:

base_model: ../evol_merge_storage/input_models/Swallow-MS-7b-v0.1_259979065
dtype: bfloat16
merge_method: dare_ties
parameters:
  int8_mask: 1.0
  normalize: 1.0
slices:
- sources:
  - layer_range: [0, 4]
    model: ../evol_merge_storage/input_models/Swallow-MS-7b-v0.1_259979065
    parameters:
      density: 1.0
      weight: 0.20736632024943663
  - layer_range: [0, 4]
    model: ../evol_merge_storage/input_models/Starling-LM-7B-beta_581094980
    parameters:
      density: 1.0
      weight: 0.2876973518761861
  - layer_range: [0, 4]
    model: ../evol_merge_storage/input_models/Mistral-7B-Instruct-v0.2_674785087
    parameters:
      density: 1.0
      weight: 0.39790911189850287
- sources:
  - layer_range: [4, 8]
    model: ../evol_merge_storage/input_models/Swallow-MS-7b-v0.1_259979065
    parameters:
      density: 1.0
      weight: 0.3259754595200053
  - layer_range: [4, 8]
    model: ../evol_merge_storage/input_models/Starling-LM-7B-beta_581094980
    parameters:
      density: 1.0
      weight: 0.36312222325553534
  - layer_range: [4, 8]
    model: ../evol_merge_storage/input_models/Mistral-7B-Instruct-v0.2_674785087
    parameters:
      density: 0.8606873476749896
      weight: 0.13151678264284256
- sources:
  - layer_range: [8, 12]
    model: ../evol_merge_storage/input_models/Swallow-MS-7b-v0.1_259979065
    parameters:
      density: 1.0
      weight: 0.16690975724594306
  - layer_range: [8, 12]
    model: ../evol_merge_storage/input_models/Starling-LM-7B-beta_581094980
    parameters:
      density: 0.8737746997323794
      weight: 0.5267457266976868
  - layer_range: [8, 12]
    model: ../evol_merge_storage/input_models/Mistral-7B-Instruct-v0.2_674785087
    parameters:
      density: 1.0
      weight: 0.37203078821341173
- sources:
  - layer_range: [12, 16]
    model: ../evol_merge_storage/input_models/Swallow-MS-7b-v0.1_259979065
    parameters:
      density: 0.9041657657943898
      weight: 0.411866096762198
  - layer_range: [12, 16]
    model: ../evol_merge_storage/input_models/Starling-LM-7B-beta_581094980
    parameters:
      density: 0.8768235480939731
      weight: 0.24309153870225503
  - layer_range: [12, 16]
    model: ../evol_merge_storage/input_models/Mistral-7B-Instruct-v0.2_674785087
    parameters:
      density: 1.0
      weight: 0.40805997159088514
- sources:
  - layer_range: [16, 20]
    model: ../evol_merge_storage/input_models/Swallow-MS-7b-v0.1_259979065
    parameters:
      density: 1.0
      weight: 0.20153807161142293
  - layer_range: [16, 20]
    model: ../evol_merge_storage/input_models/Starling-LM-7B-beta_581094980
    parameters:
      density: 1.0
      weight: 0.2651496946837373
  - layer_range: [16, 20]
    model: ../evol_merge_storage/input_models/Mistral-7B-Instruct-v0.2_674785087
    parameters:
      density: 0.881089793974409
      weight: 0.018551645245409754
- sources:
  - layer_range: [20, 24]
    model: ../evol_merge_storage/input_models/Swallow-MS-7b-v0.1_259979065
    parameters:
      density: 1.0
      weight: 0.05396099731564888
  - layer_range: [20, 24]
    model: ../evol_merge_storage/input_models/Starling-LM-7B-beta_581094980
    parameters:
      density: 1.0
      weight: 0.2544355076223701
  - layer_range: [20, 24]
    model: ../evol_merge_storage/input_models/Mistral-7B-Instruct-v0.2_674785087
    parameters:
      density: 1.0
      weight: 0.17428773365086464
- sources:
  - layer_range: [24, 28]
    model: ../evol_merge_storage/input_models/Swallow-MS-7b-v0.1_259979065
    parameters:
      density: 0.9948454730348346
      weight: 0.13561950438761128
  - layer_range: [24, 28]
    model: ../evol_merge_storage/input_models/Starling-LM-7B-beta_581094980
    parameters:
      density: 0.9012771361348846
      weight: 0.21474768477949524
  - layer_range: [24, 28]
    model: ../evol_merge_storage/input_models/Mistral-7B-Instruct-v0.2_674785087
    parameters:
      density: 0.5686565104560466
      weight: 0.5862075607169237
- sources:
  - layer_range: [28, 32]
    model: ../evol_merge_storage/input_models/Swallow-MS-7b-v0.1_259979065
    parameters:
      density: 0.7293804704051091
      weight: 0.5832263789977623
  - layer_range: [28, 32]
    model: ../evol_merge_storage/input_models/Starling-LM-7B-beta_581094980
    parameters:
      density: 1.0
      weight: 0.25251733788362796
  - layer_range: [28, 32]
    model: ../evol_merge_storage/input_models/Mistral-7B-Instruct-v0.2_674785087
    parameters:
      density: 1.0
      weight: 0.7295319486730514
tokenizer_source: base

Downloads last month: 2

Safetensors

Model size

7B params

Tensor type

BF16

Papers for HachiML/Swallow-MS-7b-MergeEvol-fevals105

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