aixk commited on
Commit
453ca45
·
1 Parent(s): fb4f383

Upload folder using huggingface_hub

Browse files
shared/global/merged/README.md ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # wwai (Transformers)
2
+
3
+ ## Purpose
4
+ - Daily conversation model
5
+ - Non-coding assistant behavior
6
+
7
+ ## Architecture
8
+ - Custom architecture: SsaiForCausalLM
9
+
10
+ - hidden_size: 1024
11
+ - intermediate_size: 4096
12
+ - layers: 16
13
+ - heads: 16
14
+ - kv_heads: 4
15
+ - max_seq_len: 160
16
+ - rope_theta: 10000
17
+ - qk_norm: True
18
+ - local_mixer: True
19
+ - neftune_alpha: 0.0
shared/global/merged/config.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "TwinyForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "dtype": "float32",
7
+ "hidden_dropout": 0.0,
8
+ "hidden_size": 1024,
9
+ "initializer_range": 0.02,
10
+ "intermediate_size": 4096,
11
+ "max_position_embeddings": 160,
12
+ "model_type": "twiny",
13
+ "neftune_alpha": 0.0,
14
+ "num_attention_heads": 16,
15
+ "num_hidden_layers": 16,
16
+ "num_key_value_heads": 4,
17
+ "qk_norm": true,
18
+ "rezero_init": 1.0,
19
+ "rms_norm_eps": 1e-06,
20
+ "rope_theta": 10000.0,
21
+ "transformers_version": "5.0.0",
22
+ "use_cache": false,
23
+ "vocab_size": 3968
24
+ }
shared/global/merged/merge_info.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"workers": ["worker-35791502"], "weights": [526.0], "merged_count": 1, "updated_at": 1776561117}
shared/global/merged/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
shared/global/merged/tokenizer_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "bos_token": "<bos>",
4
+ "clean_up_tokenization_spaces": false,
5
+ "eos_token": "<eos>",
6
+ "extra_special_tokens": [
7
+ "<|im_start|>",
8
+ "<|im_end|>"
9
+ ],
10
+ "is_local": true,
11
+ "model_max_length": 1000000,
12
+ "pad_token": "<pad>",
13
+ "tokenizer_class": "TokenizersBackend",
14
+ "unk_token": "<unk>"
15
+ }