Resolving Interference When Merging Models
Paper • 2306.01708 • Published • 19
How to use mllm-dev/gpt2_m_experiment_drug_data_ties_test with Transformers:
# Use a pipeline as a high-level helper
from transformers import pipeline
pipe = pipeline("text-classification", model="mllm-dev/gpt2_m_experiment_drug_data_ties_test") # Load model directly
from transformers import AutoTokenizer, AutoModelForSequenceClassification
tokenizer = AutoTokenizer.from_pretrained("mllm-dev/gpt2_m_experiment_drug_data_ties_test")
model = AutoModelForSequenceClassification.from_pretrained("mllm-dev/gpt2_m_experiment_drug_data_ties_test")This is a merge of pre-trained language models created using mergekit.
This model was merged using the TIES merge method using mllm-dev/gpt2_f_experiment_0_drug_data as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
base_model:
model:
path: mllm-dev/gpt2_f_experiment_0_drug_data
dtype: float16
merge_method: ties
parameters:
normalize: 1.0
slices:
- sources:
- layer_range: [0, 12]
model:
model:
path: mllm-dev/gpt2_f_experiment_0_drug_data
- layer_range: [0, 12]
model:
model:
path: mllm-dev/gpt2_f_experiment_1_drug_data
parameters:
density: 0.9
weight: 0.2
- layer_range: [0, 12]
model:
model:
path: mllm-dev/gpt2_f_experiment_2_drug_data
parameters:
density: 0.9
weight: 0.2
- layer_range: [0, 12]
model:
model:
path: mllm-dev/gpt2_f_experiment_3_drug_data
parameters:
density: 0.9
weight: 0.2
- layer_range: [0, 12]
model:
model:
path: mllm-dev/gpt2_f_experiment_4_drug_data
parameters:
density: 0.9
weight: 0.2