Neo_7b-merge16

Neo_7b-merge16 is a merge of the following models using LazyMergekit:

• DewEfresh/neo_7b
• DewEfresh/neo_7b

🧩 Configuration

# Define the slices for the model merging process
slices:
  - sources:
      # Merge layer 3 with layer 0
      - model: DewEfresh/neo_7b
        layer_range: [3, 3]
      - model: DewEfresh/neo_7b
        layer_range: [0, 0]
  - sources:
      # Merge layer 3 with layer 1
      - model: DewEfresh/neo_7b
        layer_range: [3, 3]
      - model: DewEfresh/neo_7b
        layer_range: [1, 1]
  - sources:
      # Merge layer 3 with layer 2
      - model: DewEfresh/neo_7b
        layer_range: [3, 3]
      - model: DewEfresh/neo_7b
        layer_range: [2, 2]
  - sources:
      # Merge layer 7 with layer 4
      - model: DewEfresh/neo_7b
        layer_range: [7, 7]
      - model: DewEfresh/neo_7b
        layer_range: [4, 4]
  - sources:
      # Merge layer 7 with layer 5
      - model: DewEfresh/neo_7b
        layer_range: [7, 7]
      - model: DewEfresh/neo_7b
        layer_range: [5, 5]
  - sources:
      # Merge layer 7 with layer 6
      - model: DewEfresh/neo_7b
        layer_range: [7, 7]
      - model: DewEfresh/neo_7b
        layer_range: [6, 6]
  - sources:
      # Merge layer 11 with layer 8
      - model: DewEfresh/neo_7b
        layer_range: [11, 11]
      - model: DewEfresh/neo_7b
        layer_range: [8, 8]
  - sources:
      # Merge layer 11 with layer 9
      - model: DewEfresh/neo_7b
        layer_range: [11, 11]
      - model: DewEfresh/neo_7b
        layer_range: [9, 9]
  - sources:
      # Merge layer 11 with layer 10
      - model: DewEfresh/neo_7b
        layer_range: [11, 11]
      - model: DewEfresh/neo_7b
        layer_range: [10, 10]
  - sources:
      # Merge layer 15 with layer 12
      - model: DewEfresh/neo_7b
        layer_range: [15, 15]
      - model: DewEfresh/neo_7b
        layer_range: [12, 12]
  - sources:
      # Merge layer 15 with layer 13
      - model: DewEfresh/neo_7b
        layer_range: [15, 15]
      - model: DewEfresh/neo_7b
        layer_range: [13, 13]
  - sources:
      # Merge layer 15 with layer 14
      - model: DewEfresh/neo_7b
        layer_range: [15, 15]
      - model: DewEfresh/neo_7b
        layer_range: [14, 14]
  - sources:
      # Merge layer 19 with layer 16
      - model: DewEfresh/neo_7b
        layer_range: [19, 19]
      - model: DewEfresh/neo_7b
        layer_range: [16, 16]
  - sources:
      # Merge layer 19 with layer 17
      - model: DewEfresh/neo_7b
        layer_range: [19, 19]
      - model: DewEfresh/neo_7b
        layer_range: [17, 17]
  - sources:
      # Merge layer 19 with layer 18
      - model: DewEfresh/neo_7b
        layer_range: [19, 19]
      - model: DewEfresh/neo_7b
        layer_range: [18, 18]
  - sources:
      # Merge layer 23 with layer 20
      - model: DewEfresh/neo_7b
        layer_range: [23, 23]
      - model: DewEfresh/neo_7b
        layer_range: [20, 20]
  - sources:
      # Merge layer 23 with layer 21
      - model: DewEfresh/neo_7b
        layer_range: [23, 23]
      - model: DewEfresh/neo_7b
        layer_range: [21, 21]
  - sources:
      # Merge layer 23 with layer 22
      - model: DewEfresh/neo_7b
        layer_range: [23, 23]
      - model: DewEfresh/neo_7b
        layer_range: [22, 22]
  - sources:
      # Merge layer 27 with layer 24
      - model: DewEfresh/neo_7b
        layer_range: [27, 27]
      - model: DewEfresh/neo_7b
        layer_range: [24, 24]
  - sources:
      # Merge layer 27 with layer 25
      - model: DewEfresh/neo_7b
        layer_range: [27, 27]
      - model: DewEfresh/neo_7b
        layer_range: [25, 25]
  - sources:
      # Merge layer 27 with layer 26
      - model: DewEfresh/neo_7b
        layer_range: [27, 27]
      - model: DewEfresh/neo_7b
        layer_range: [26, 26]
# Specify the merging method for the slices
merge_method: slerp
base_model: DewEfresh/neo_7b
parameters:
  t: 0.3333  # Set global interpolation value to 33.33%
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "DewEfresh/Neo_7b-merge16"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])