Hugging Face's logo

basically-ai
/
Pebble-25M

Text Generation
Transformers
Safetensors
PyTorch
English
pebble_25m
pebble
language-model
base-model
small-language-model
custom-code
mamba2
hybrid
custom_code
Model card Files
xet
 Community
Hoglet-33 commited on
Commit
2d607b1
·
verified ·
1 Parent(s): d718e7b

Update config.json

Browse files
Files changed (1) hide show
  1. config.json +27 -17
config.json CHANGED
@@ -1,22 +1,32 @@
1
  {
2
- "model_type": "pebble_25m",
3
  "architectures": [
4
- "PebbleLM"
5
  ],
 
 
 
 
 
6
  "vocab_size": 2048,
7
- "d_model": 608,
8
- "n_blocks": 8,
9
- "n_heads": 8,
10
- "att_hidden": 2432,
11
- "rope_theta": 10000.0,
12
- "seq_len": 2048,
13
- "eos_token_id": 0,
14
- "bos_token_id": 0,
15
- "torch_dtype": "float32",
16
- "mamba_d_state": 128,
17
- "mamba_d_conv": 4,
18
- "mamba_expand": 2,
19
- "mamba_headdim": 64,
20
- "block_pattern": "3:1 mamba:attn",
21
- "transformers_version": "4.44.0"
 
 
 
 
 
 
22
  }
 
1
  {
 
2
  "architectures": [
3
+ "PebbleForCausalLM"
4
  ],
5
+ "model_type": "pebble_25m",
6
+ "auto_map": {
7
+ "AutoConfig": "configuration_pebble.PebbleConfig",
8
+ "AutoModelForCausalLM": "modeling_pebble.PebbleForCausalLM"
9
+ },
10
  "vocab_size": 2048,
11
+ "hidden_size": 608,
12
+ "intermediate_size": 2432,
13
+ "num_hidden_layers": 8,
14
+ "num_attention_heads": 8,
15
+ "block_pattern": "mmma|mmma",
16
+ "hybrid_ratio": "3:1 mamba2:attention",
17
+ "max_position_embeddings": 2048,
18
+ "rms_norm_eps": 1e-06,
19
+ "tie_word_embeddings": true,
20
+ "mamba2": {
21
+ "d_state": 128,
22
+ "d_conv": 4,
23
+ "expand": 2,
24
+ "headdim": 64,
25
+ "use_mem_eff_path": true
26
+ },
27
+ "attention": {
28
+ "rope_theta": 10000.0,
29
+ "is_causal": true
30
+ },
31
+ "torch_dtype": "float32"
32
  }