tiny-random/baguettotron

This tiny model is intended for debugging. It is randomly initialized using the configuration adapted from PleIAs/Baguettotron.

Example usage:

from transformers import pipeline
model_id = "tiny-random/baguettotron"
pipe = pipeline(
    "text-generation", model=model_id, device="cuda",
    trust_remote_code=True, max_new_tokens=3,
)
print(pipe("Hello World!"))

Codes to create this repo:

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

source_model_id = "PleIAs/Baguettotron"
save_folder = "/tmp/tiny-random/baguettotron"

tokenizer = AutoTokenizer.from_pretrained(
    source_model_id, trust_remote_code=True,
)
tokenizer.chat_template = "{% for m in messages %}<|im_start|>{{ m['role'] }}\n{{ m['content'] }}<|im_end|>\n{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant\n<think>\n{% endif %}"
tokenizer.eos_token = "<|im_end|>"
tokenizer.bos_token = "<|im_start|>"
tokenizer.stop_token = "<|im_end|>"
tokenizer.save_pretrained(save_folder)

config = AutoConfig.from_pretrained(
    source_model_id, trust_remote_code=True,
)
config.hidden_size = 8
config.intermediate_size = 64
config.num_attention_heads = 16
config.num_key_value_heads = 8
config.head_dim = 32
config.num_hidden_layers = 2

model = AutoModelForCausalLM.from_config(
    config,
    torch_dtype=torch.bfloat16,
    trust_remote_code=True,
)
model.generation_config = GenerationConfig.from_pretrained(
    source_model_id, trust_remote_code=True,
)
set_seed(42)
model = model.cpu()
with torch.no_grad():
    for name, p in sorted(model.named_parameters()):
        torch.nn.init.normal_(p, 0, 0.1)
        print(name, p.shape)
model.save_pretrained(save_folder)

Printing the model:

LlamaForCausalLM(
  (model): LlamaModel(
    (embed_tokens): Embedding(65536, 8)
    (layers): ModuleList(
      (0-1): 2 x LlamaDecoderLayer(
        (self_attn): LlamaAttention(
          (q_proj): Linear(in_features=8, out_features=512, bias=False)
          (k_proj): Linear(in_features=8, out_features=256, bias=False)
          (v_proj): Linear(in_features=8, out_features=256, bias=False)
          (o_proj): Linear(in_features=512, out_features=8, bias=False)
        )
        (mlp): LlamaMLP(
          (gate_proj): Linear(in_features=8, out_features=64, bias=False)
          (up_proj): Linear(in_features=8, out_features=64, bias=False)
          (down_proj): Linear(in_features=64, out_features=8, bias=False)
          (act_fn): SiLUActivation()
        )
        (input_layernorm): LlamaRMSNorm((8,), eps=1e-05)
        (post_attention_layernorm): LlamaRMSNorm((8,), eps=1e-05)
      )
    )
    (norm): LlamaRMSNorm((8,), eps=1e-05)
    (rotary_emb): LlamaRotaryEmbedding()
  )
  (lm_head): Linear(in_features=8, out_features=65536, bias=False)
)