metadata

tags:
  - merge
  - mergekit
  - lazymergekit
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
base_model:
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B
  - yam-peleg/Experiment26-7B

Experiment26-7B-passthrough-10slice

Experiment26-7B-passthrough-10slice is a merge of the following models using LazyMergekit:

🧩 Configuration

slices:
  - sources:
    - model: yam-peleg/Experiment26-7B
      layer_range: [0, 5]
  - sources:
    - model: yam-peleg/Experiment26-7B
      layer_range: [3, 8]
  - sources:
    - model: yam-peleg/Experiment26-7B
      layer_range: [6, 11]
  - sources:
    - model: yam-peleg/Experiment26-7B
      layer_range: [9, 14]
  - sources:
    - model: yam-peleg/Experiment26-7B
      layer_range: [12, 17]
  - sources:
    - model: yam-peleg/Experiment26-7B
      layer_range: [15, 20]
  - sources:
    - model: yam-peleg/Experiment26-7B
      layer_range: [18, 23]
  - sources:
    - model: yam-peleg/Experiment26-7B
      layer_range: [21, 26]
  - sources:
    - model: yam-peleg/Experiment26-7B
      layer_range: [24, 29]
  - sources:
    - model: yam-peleg/Experiment26-7B
      layer_range: [27, 32]
merge_method: passthrough
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "babybirdprd/Experiment26-7B-passthrough-10slice"
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"])