111111 commited on
Commit
db9833b
·
verified ·
1 Parent(s): a79f99b

Trained with Unsloth - config

Browse files
Files changed (1) hide show
  1. config.json +67 -0
config.json ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Lfm2ForCausalLM"
4
+ ],
5
+ "block__name_mlp": "parallel_mlp_merged",
6
+ "block_auto_adjust_ff_dim": false,
7
+ "block_dim": 1024,
8
+ "block_ffn_dim_multiplier": 1.0,
9
+ "block_ffn_te_autocast": false,
10
+ "block_ffn_use_quantized_params": false,
11
+ "block_mlp_init_scale": 1.0,
12
+ "block_multiple_of": 256,
13
+ "block_norm_eps": 1e-05,
14
+ "block_out_init_scale": 1.0,
15
+ "block_sequence_parallel_norm_across_tp": false,
16
+ "block_use_swiglu": true,
17
+ "block_use_xavier_init": true,
18
+ "bos_token_id": 1,
19
+ "conv_L_cache": 3,
20
+ "conv_bias": false,
21
+ "conv_dim": 1024,
22
+ "conv_use_xavier_init": true,
23
+ "dtype": "bfloat16",
24
+ "eos_token_id": 7,
25
+ "ffn_te_autocast": false,
26
+ "ffn_use_quantized_params": false,
27
+ "hidden_size": 1024,
28
+ "initializer_range": 0.02,
29
+ "intermediate_size": 2560,
30
+ "layer_types": [
31
+ "conv",
32
+ "conv",
33
+ "full_attention",
34
+ "conv",
35
+ "full_attention",
36
+ "conv",
37
+ "full_attention",
38
+ "conv",
39
+ "full_attention",
40
+ "conv",
41
+ "full_attention",
42
+ "conv",
43
+ "full_attention",
44
+ "conv"
45
+ ],
46
+ "max_position_embeddings": 128000,
47
+ "model_name": "unsloth/LFM2.5-230M",
48
+ "model_type": "lfm2",
49
+ "norm_eps": 1e-05,
50
+ "num_attention_heads": 16,
51
+ "num_heads": 16,
52
+ "num_hidden_layers": 14,
53
+ "num_key_value_heads": 8,
54
+ "pad_token_id": 0,
55
+ "rope_parameters": {
56
+ "rope_theta": 1000000.0,
57
+ "rope_type": "default"
58
+ },
59
+ "sequence_parallel_norm_across_tp": false,
60
+ "tie_word_embeddings": true,
61
+ "transformers_version": "5.3.0",
62
+ "unsloth_fixed": true,
63
+ "unsloth_version": "2026.7.2",
64
+ "use_cache": true,
65
+ "use_pos_enc": true,
66
+ "vocab_size": 65536
67
+ }