This model is randomly initialized, using the config from [https://huggingface.co/tiiuae/falcon-7b-instruct] but with smaller size. Note the model is in float16.
Codes:
import transformers
from optimum.intel.openvino import OVModelForCausalLM
import torch
import os
from huggingface_hub import create_repo, upload_folder
source_model_id = 'tiiuae/falcon-7b-instruct'
save_path = '/tmp/yujiepan/falcon-tiny-random'
repo_id = 'yujiepan/falcon-tiny-random'
config = transformers.AutoConfig.from_pretrained(
source_model_id, trust_remote_code=True)
config.hidden_size = 8
config.num_attention_heads = 2
config.num_hidden_layers = 2
config.torch_dtype = torch.float16
model = transformers.AutoModelForCausalLM.from_config(
config, trust_remote_code=True)
model = model.half()
model.save_pretrained(save_path)
tokenizer = transformers.AutoTokenizer.from_pretrained(
source_model_id, trust_remote_code=True)
tokenizer.save_pretrained(save_path)
# current not supported, might add this later
# ovmodel = OVModelForCausalLM.from_pretrained(
# save_path, export=True, trust_remote_code=True)
# ovmodel.save_pretrained(save_path)
os.system(f'ls -alh {save_path}')
create_repo(repo_id, exist_ok=True)
upload_folder(repo_id=repo_id, folder_path=save_path)