Upload folder using huggingface_hub
Browse files- README.md +19 -0
- config.json +24 -0
- tokenizer.json +0 -0
- tokenizer_config.json +15 -0
README.md
ADDED
|
@@ -0,0 +1,19 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# axai-stack-0_3B (Transformers)
|
| 2 |
+
|
| 3 |
+
## Purpose
|
| 4 |
+
- Daily conversation model
|
| 5 |
+
- Non-coding assistant behavior
|
| 6 |
+
|
| 7 |
+
## Architecture
|
| 8 |
+
- Custom architecture: AxaiForCausalLM
|
| 9 |
+
|
| 10 |
+
- hidden_size: 768
|
| 11 |
+
- intermediate_size: 3072
|
| 12 |
+
- layers: 24
|
| 13 |
+
- heads: 12
|
| 14 |
+
- kv_heads: 6
|
| 15 |
+
- max_seq_len: 128
|
| 16 |
+
- rope_theta: 10000
|
| 17 |
+
- qk_norm: True
|
| 18 |
+
- local_mixer: True
|
| 19 |
+
- neftune_alpha: 0.0
|
config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"AxaiForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_dropout": 0.0,
|
| 6 |
+
"dtype": "float32",
|
| 7 |
+
"hidden_dropout": 0.0,
|
| 8 |
+
"hidden_size": 768,
|
| 9 |
+
"initializer_range": 0.02,
|
| 10 |
+
"intermediate_size": 3072,
|
| 11 |
+
"max_position_embeddings": 128,
|
| 12 |
+
"model_type": "axai",
|
| 13 |
+
"neftune_alpha": 0.0,
|
| 14 |
+
"num_attention_heads": 12,
|
| 15 |
+
"num_hidden_layers": 24,
|
| 16 |
+
"num_key_value_heads": 6,
|
| 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": 32000
|
| 24 |
+
}
|
tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
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 |
+
}
|