README.md

---
library_name: transformers
pipeline_tag: text-generation
inference: true
widget:
- text: Hello!
  example_title: Hello world
  group: Python
---

This model is randomly initialized, using the config from [THUDM/glm-4-9b-chat](https://huggingface.co/THUDM/glm-4-9b-chat) but with smaller size. 

Codes:

```python
import transformers
import torch
import os
from huggingface_hub import create_repo, upload_folder
import accelerate

source_model_id = 'THUDM/glm-4-9b-chat'
save_path = '/tmp/yujiepan/glm-4-tiny-random'
repo_id = 'yujiepan/glm-4-tiny-random'

os.system(f'rm -rf {save_path}')

config = transformers.AutoConfig.from_pretrained(
    source_model_id,
    trust_remote_code=True,
)
config._name_or_path = source_model_id
config.hidden_size = 8
config.ffn_hidden_size = 16
config.kv_channels = 2
config.num_attention_heads = 4
config.multi_query_group_num = 2
config.num_hidden_layers = 2
config.num_layers = 2

model = transformers.AutoModelForCausalLM.from_config(
    config,
    trust_remote_code=True,
)
model.generation_config = transformers.GenerationConfig.from_pretrained(source_model_id)
model = model.to(torch.bfloat16)

with torch.no_grad():
    for p in model.parameters():
        torch.nn.init.normal_(p)

model.save_pretrained(save_path)

tokenizer = transformers.AutoTokenizer.from_pretrained(
    source_model_id,
    trust_remote_code=True,
)
tokenizer.save_pretrained(save_path)

output = model.float().generate(torch.tensor([[1, 2, 3]]).long(), max_length=16, do_sample=True)

os.system(f'ls -alh {save_path}')
# os.system(f'rm -rf {save_path}/model.safetensors')
create_repo(repo_id, exist_ok=True)
upload_folder(repo_id=repo_id, folder_path=save_path)
```