File size: 2,839 Bytes
2c87f05
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
---
library_name: transformers
pipeline_tag: text-generation
inference: true
widget:
  - text: Hello!
    example_title: Hello world
    group: Python
---

This tiny model is for debugging. It is randomly initialized with the config adapted from [meta-llama/Llama-3.3-70B-Instruct](https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct).

### Example usage:

```python
from transformers import pipeline
model_id = "tiny-random/llama-3.3-dim64"
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:

```python
import torch

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

source_model_id = "meta-llama/Llama-3.3-70B-Instruct"
save_folder = "/tmp/tiny-random/llama-3.3-dim64"

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

config = AutoConfig.from_pretrained(
    source_model_id, trust_remote_code=True,
)
config.hidden_size = 64
config.intermediate_size = 128
config.num_attention_heads = 2
config.num_key_value_heads = 1
config.head_dim = 32
config.num_hidden_layers = 2
config.tie_word_embeddings = True

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)
with torch.no_grad():
    for name, p in sorted(model.named_parameters()):
        torch.nn.init.normal_(p, 0, 0.2)
        print(name, p.shape)
model.save_pretrained(save_folder)
```

### Printing the model:

```text
LlamaForCausalLM(
  (model): LlamaModel(
    (embed_tokens): Embedding(128256, 64)
    (layers): ModuleList(
      (0-1): 2 x LlamaDecoderLayer(
        (self_attn): LlamaAttention(
          (q_proj): Linear(in_features=64, out_features=64, bias=False)
          (k_proj): Linear(in_features=64, out_features=32, bias=False)
          (v_proj): Linear(in_features=64, out_features=32, bias=False)
          (o_proj): Linear(in_features=64, out_features=64, bias=False)
        )
        (mlp): LlamaMLP(
          (gate_proj): Linear(in_features=64, out_features=128, bias=False)
          (up_proj): Linear(in_features=64, out_features=128, bias=False)
          (down_proj): Linear(in_features=128, out_features=64, bias=False)
          (act_fn): SiLU()
        )
        (input_layernorm): LlamaRMSNorm((64,), eps=1e-05)
        (post_attention_layernorm): LlamaRMSNorm((64,), eps=1e-05)
      )
    )
    (norm): LlamaRMSNorm((64,), eps=1e-05)
    (rotary_emb): LlamaRotaryEmbedding()
  )
  (lm_head): Linear(in_features=64, out_features=128256, bias=False)
)
```