metadata
pipeline_tag: text-generation
inference: true
widget:
- text: Hello!
example_title: Hello world
group: Python
library_name: transformers
This model is randomly initialized, using the config from tiiuae/falcon-40b but with smaller size. Note:
- The model uses "new architecture" in Falcon-40b.
- 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-40b'
save_path = '/tmp/yujiepan/falcon-new-tiny64-random'
repo_id = 'yujiepan/falcon-new-tiny64-random'
config = transformers.AutoConfig.from_pretrained(
source_model_id, trust_remote_code=True)
config.hidden_size = 64
config.num_attention_heads = 2
config.num_hidden_layers = 1
config.num_kv_heads = 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)