README.md

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

This model is for debugging. It is randomly initialized with the config from [ibm-fms/Bamba-9B](https://huggingface.co/ibm-fms/Bamba-9B) but is of smaller size. 

Codes:
```python
import os

import torch
import transformers
from transformers import (AutoConfig, AutoModelForCausalLM, AutoTokenizer,
                          GenerationConfig, pipeline, set_seed)

model_id = "ibm-fms/Bamba-9B"
repo_id = "tiny-random/bamba"
save_path = f"/tmp/{repo_id}"

config = AutoConfig.from_pretrained(model_id, trust_remote_code=True)
config.attn_layer_indices = [1]
config.attn_rotary_emb = 4
config.hidden_size = 16
config.intermediate_size = 32
config.num_attention_heads = 2
config.num_hidden_layers = 2
config.num_key_value_heads = 1
config.mamba_expand = 4
config.mamba_d_head = 8
config.mamba_n_heads = 8
config.mamba_d_state = 8

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

model = AutoModelForCausalLM.from_config(
    config, torch_dtype=torch.bfloat16,
    trust_remote_code=True,
)
# model.generation_config = GenerationConfig.from_pretrained(
#     model_id, trust_remote_code=True
# )


set_seed(42)
with torch.no_grad():
    for name, p in sorted(model.named_parameters()):
        torch.nn.init.normal_(p, 0, 0.5)
        print(name, p.shape)

model.save_pretrained(save_path)

model = AutoModelForCausalLM.from_pretrained(save_path).cuda()
tokenizer = AutoTokenizer.from_pretrained(save_path)
message = ["Hello, world!"]
inputs = tokenizer(message, return_tensors='pt', return_token_type_ids=False).to(model.device)
response = model.generate(**inputs, max_new_tokens=2)[0]
print(tokenizer.convert_ids_to_tokens(response, skip_special_tokens=False))
```