Make model tiny (172M -> 4.5M params)

#2
by lewtun HF staff - opened
Files changed (2) hide show
  1. config.json +8 -6
  2. pytorch_model.bin +2 -2
config.json CHANGED
@@ -1,11 +1,13 @@
1
  {
2
- "_name_or_path": "google/mt5-small",
3
  "architectures": [
4
  "MT5Model"
5
  ],
6
  "d_ff": 1024,
7
- "d_kv": 64,
8
- "d_model": 512,
 
 
9
  "decoder_start_token_id": 0,
10
  "dense_act_fn": "gelu_new",
11
  "dropout_rate": 0.1,
@@ -17,15 +19,15 @@
17
  "layer_norm_epsilon": 1e-06,
18
  "model_type": "mt5",
19
  "num_decoder_layers": 8,
20
- "num_heads": 6,
21
- "num_layers": 8,
22
  "pad_token_id": 0,
23
  "relative_attention_max_distance": 128,
24
  "relative_attention_num_buckets": 32,
25
  "tie_word_embeddings": false,
26
  "tokenizer_class": "T5Tokenizer",
27
  "torch_dtype": "float32",
28
- "transformers_version": "4.21.1",
29
  "use_cache": true,
30
  "vocab_size": 250112
31
  }
 
1
  {
2
+ "_name_or_path": "hf-internal-testing/tiny-random-mt5",
3
  "architectures": [
4
  "MT5Model"
5
  ],
6
  "d_ff": 1024,
7
+ "d_kv": 4,
8
+ "d_model": 16,
9
+ "decoder_attention_heads": 4,
10
+ "decoder_ffn_dim": 4,
11
  "decoder_start_token_id": 0,
12
  "dense_act_fn": "gelu_new",
13
  "dropout_rate": 0.1,
 
19
  "layer_norm_epsilon": 1e-06,
20
  "model_type": "mt5",
21
  "num_decoder_layers": 8,
22
+ "num_heads": 4,
23
+ "num_layers": 2,
24
  "pad_token_id": 0,
25
  "relative_attention_max_distance": 128,
26
  "relative_attention_num_buckets": 32,
27
  "tie_word_embeddings": false,
28
  "tokenizer_class": "T5Tokenizer",
29
  "torch_dtype": "float32",
30
+ "transformers_version": "4.22.0.dev0",
31
  "use_cache": true,
32
  "vocab_size": 250112
33
  }
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3c5b7434d0b5e2ea1766572e8c70f235239199c0581cc5339b0a9238e80cb97f
3
- size 688538311
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:728f60112bd2377b5ff8a823d8f8b1d0b18283ca8aae6b171590e81c1707b213
3
+ size 18092403