Hugging Face's logo

LiquidAI
/
LFM2.5-350M-MLX-bf16

Text Generation
MLX
Safetensors
lfm2
liquid
lfm2.5
edge
conversational
Model card Files
xet
 Community
ykhrustalev commited on
Commit
72556de
·
verified ·
1 Parent(s): 582249a

Upload folder using huggingface_hub

Browse files
Files changed (7) hide show
  1. README.md +82 -0
  2. config.json +62 -0
  3. generation_config.json +7 -0
  4. model.safetensors +3 -0
  5. model.safetensors.index.json +156 -0
  6. tokenizer.json +0 -0
  7. tokenizer_config.json +19 -0
README.md ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: mlx
3
+ license: other
4
+ license_name: lfm1.0
5
+ license_link: LICENSE
6
+ language:
7
+ - en
8
+ - ar
9
+ - zh
10
+ - fr
11
+ - de
12
+ - ja
13
+ - ko
14
+ - es
15
+ pipeline_tag: text-generation
16
+ tags:
17
+ - liquid
18
+ - lfm2.5
19
+ - edge
20
+ - mlx
21
+ base_model: LiquidAI/LFM2.5-350M
22
+ ---
23
+
24
+ <div align="center">
25
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/2b08LKpev0DNEk6DlnWkY.png" alt="Liquid AI" style="width: 100%; max-width: 100%;">
26
+
27
+ <p>
28
+ <a href="https://playground.liquid.ai/"><strong>Try LFM</strong></a> •
29
+ <a href="https://docs.liquid.ai/lfm"><strong>Documentation</strong></a> •
30
+ <a href="https://leap.liquid.ai/"><strong>LEAP</strong></a> •
31
+ <a href="https://www.liquid.ai/blog/"><strong>Blog</strong></a>
32
+ </p>
33
+ </div>
34
+
35
+ # LFM2.5-350M-MLX-bf16
36
+
37
+ MLX export of [LFM2.5-350M](https://huggingface.co/LiquidAI/LFM2.5-350M) for Apple Silicon inference.
38
+
39
+ LFM2.5-350M is a compact multilingual base model built on LiquidAI's hybrid architecture, combining convolutional and attention layers for efficient long-context processing.
40
+
41
+ ## Model Details
42
+
43
+ | Property | Value |
44
+ |----------|-------|
45
+ | Parameters | 350M |
46
+ | Precision | bf16 |
47
+ | Size | 676 MB |
48
+ | Context Length | 128K |
49
+
50
+ ## Use with mlx
51
+
52
+ ```bash
53
+ pip install mlx-lm
54
+ ```
55
+
56
+ ```python
57
+ from mlx_lm import load, generate
58
+ from mlx_lm.sample_utils import make_sampler
59
+
60
+ model, tokenizer = load("LiquidAI/LFM2.5-350M-MLX-bf16")
61
+
62
+ response = generate(
63
+ model,
64
+ tokenizer,
65
+ prompt="The capital of France is",
66
+ max_tokens=100,
67
+ sampler=make_sampler(temp=0.7),
68
+ verbose=True,
69
+ )
70
+ ```
71
+
72
+ ## Other Precisions
73
+
74
+ - [LFM2.5-350M-MLX-bf16](https://huggingface.co/LiquidAI/LFM2.5-350M-MLX-bf16) (676 MB)
75
+ - [LFM2.5-350M-MLX-8bit](https://huggingface.co/LiquidAI/LFM2.5-350M-MLX-8bit) (381 MB)
76
+ - [LFM2.5-350M-MLX-6bit](https://huggingface.co/LiquidAI/LFM2.5-350M-MLX-6bit) (296 MB)
77
+ - [LFM2.5-350M-MLX-5bit](https://huggingface.co/LiquidAI/LFM2.5-350M-MLX-5bit) (254 MB)
78
+ - [LFM2.5-350M-MLX-4bit](https://huggingface.co/LiquidAI/LFM2.5-350M-MLX-4bit) (212 MB)
79
+
80
+ ## License
81
+
82
+ This model is released under the [LFM 1.0 License](LICENSE).
config.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Lfm2ForCausalLM"
4
+ ],
5
+ "attn_implementation": "flash_attention_2",
6
+ "block_auto_adjust_ff_dim": true,
7
+ "block_dim": 1024,
8
+ "block_ff_dim": 6656,
9
+ "block_ffn_dim_multiplier": 1.0,
10
+ "block_mlp_init_scale": 1.0,
11
+ "block_multiple_of": 256,
12
+ "block_norm_eps": 1e-05,
13
+ "block_out_init_scale": 1.0,
14
+ "block_use_swiglu": true,
15
+ "block_use_xavier_init": true,
16
+ "bos_token_id": 1,
17
+ "conv_L_cache": 3,
18
+ "conv_bias": false,
19
+ "conv_dim": 1024,
20
+ "conv_use_xavier_init": true,
21
+ "dtype": "bfloat16",
22
+ "eos_token_id": 7,
23
+ "hidden_size": 1024,
24
+ "initializer_range": 0.02,
25
+ "intermediate_size": 6656,
26
+ "layer_types": [
27
+ "conv",
28
+ "conv",
29
+ "full_attention",
30
+ "conv",
31
+ "conv",
32
+ "full_attention",
33
+ "conv",
34
+ "conv",
35
+ "full_attention",
36
+ "conv",
37
+ "full_attention",
38
+ "conv",
39
+ "full_attention",
40
+ "conv",
41
+ "full_attention",
42
+ "conv"
43
+ ],
44
+ "max_position_embeddings": 128000,
45
+ "model_type": "lfm2",
46
+ "norm_eps": 1e-05,
47
+ "num_attention_heads": 16,
48
+ "num_heads": 16,
49
+ "num_hidden_layers": 16,
50
+ "num_key_value_heads": 8,
51
+ "pad_token_id": 0,
52
+ "rope_parameters": {
53
+ "rope_theta": 1000000.0,
54
+ "rope_type": "default"
55
+ },
56
+ "rope_theta": 1000000.0,
57
+ "tie_embedding": true,
58
+ "transformers_version": "4.57.1",
59
+ "use_cache": true,
60
+ "use_pos_enc": true,
61
+ "vocab_size": 65536
62
+ }
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 7,
5
+ "pad_token_id": 0,
6
+ "transformers_version": "4.57.1"
7
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97eb6787ed857d49c5b9b692a6fa8d2baeaf8582bca128b08aa19b389ba9cee0
3
+ size 708984390
model.safetensors.index.json ADDED
@@ -0,0 +1,156 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 708967936,
4
+ "total_parameters": 354483968
5
+ },
6
+ "weight_map": {
7
+ "model.embed_tokens.weight": "model.safetensors",
8
+ "model.embedding_norm.weight": "model.safetensors",
9
+ "model.layers.0.conv.conv.weight": "model.safetensors",
10
+ "model.layers.0.conv.in_proj.weight": "model.safetensors",
11
+ "model.layers.0.conv.out_proj.weight": "model.safetensors",
12
+ "model.layers.0.feed_forward.w1.weight": "model.safetensors",
13
+ "model.layers.0.feed_forward.w2.weight": "model.safetensors",
14
+ "model.layers.0.feed_forward.w3.weight": "model.safetensors",
15
+ "model.layers.0.ffn_norm.weight": "model.safetensors",
16
+ "model.layers.0.operator_norm.weight": "model.safetensors",
17
+ "model.layers.1.conv.conv.weight": "model.safetensors",
18
+ "model.layers.1.conv.in_proj.weight": "model.safetensors",
19
+ "model.layers.1.conv.out_proj.weight": "model.safetensors",
20
+ "model.layers.1.feed_forward.w1.weight": "model.safetensors",
21
+ "model.layers.1.feed_forward.w2.weight": "model.safetensors",
22
+ "model.layers.1.feed_forward.w3.weight": "model.safetensors",
23
+ "model.layers.1.ffn_norm.weight": "model.safetensors",
24
+ "model.layers.1.operator_norm.weight": "model.safetensors",
25
+ "model.layers.10.feed_forward.w1.weight": "model.safetensors",
26
+ "model.layers.10.feed_forward.w2.weight": "model.safetensors",
27
+ "model.layers.10.feed_forward.w3.weight": "model.safetensors",
28
+ "model.layers.10.ffn_norm.weight": "model.safetensors",
29
+ "model.layers.10.operator_norm.weight": "model.safetensors",
30
+ "model.layers.10.self_attn.k_layernorm.weight": "model.safetensors",
31
+ "model.layers.10.self_attn.k_proj.weight": "model.safetensors",
32
+ "model.layers.10.self_attn.out_proj.weight": "model.safetensors",
33
+ "model.layers.10.self_attn.q_layernorm.weight": "model.safetensors",
34
+ "model.layers.10.self_attn.q_proj.weight": "model.safetensors",
35
+ "model.layers.10.self_attn.v_proj.weight": "model.safetensors",
36
+ "model.layers.11.conv.conv.weight": "model.safetensors",
37
+ "model.layers.11.conv.in_proj.weight": "model.safetensors",
38
+ "model.layers.11.conv.out_proj.weight": "model.safetensors",
39
+ "model.layers.11.feed_forward.w1.weight": "model.safetensors",
40
+ "model.layers.11.feed_forward.w2.weight": "model.safetensors",
41
+ "model.layers.11.feed_forward.w3.weight": "model.safetensors",
42
+ "model.layers.11.ffn_norm.weight": "model.safetensors",
43
+ "model.layers.11.operator_norm.weight": "model.safetensors",
44
+ "model.layers.12.feed_forward.w1.weight": "model.safetensors",
45
+ "model.layers.12.feed_forward.w2.weight": "model.safetensors",
46
+ "model.layers.12.feed_forward.w3.weight": "model.safetensors",
47
+ "model.layers.12.ffn_norm.weight": "model.safetensors",
48
+ "model.layers.12.operator_norm.weight": "model.safetensors",
49
+ "model.layers.12.self_attn.k_layernorm.weight": "model.safetensors",
50
+ "model.layers.12.self_attn.k_proj.weight": "model.safetensors",
51
+ "model.layers.12.self_attn.out_proj.weight": "model.safetensors",
52
+ "model.layers.12.self_attn.q_layernorm.weight": "model.safetensors",
53
+ "model.layers.12.self_attn.q_proj.weight": "model.safetensors",
54
+ "model.layers.12.self_attn.v_proj.weight": "model.safetensors",
55
+ "model.layers.13.conv.conv.weight": "model.safetensors",
56
+ "model.layers.13.conv.in_proj.weight": "model.safetensors",
57
+ "model.layers.13.conv.out_proj.weight": "model.safetensors",
58
+ "model.layers.13.feed_forward.w1.weight": "model.safetensors",
59
+ "model.layers.13.feed_forward.w2.weight": "model.safetensors",
60
+ "model.layers.13.feed_forward.w3.weight": "model.safetensors",
61
+ "model.layers.13.ffn_norm.weight": "model.safetensors",
62
+ "model.layers.13.operator_norm.weight": "model.safetensors",
63
+ "model.layers.14.feed_forward.w1.weight": "model.safetensors",
64
+ "model.layers.14.feed_forward.w2.weight": "model.safetensors",
65
+ "model.layers.14.feed_forward.w3.weight": "model.safetensors",
66
+ "model.layers.14.ffn_norm.weight": "model.safetensors",
67
+ "model.layers.14.operator_norm.weight": "model.safetensors",
68
+ "model.layers.14.self_attn.k_layernorm.weight": "model.safetensors",
69
+ "model.layers.14.self_attn.k_proj.weight": "model.safetensors",
70
+ "model.layers.14.self_attn.out_proj.weight": "model.safetensors",
71
+ "model.layers.14.self_attn.q_layernorm.weight": "model.safetensors",
72
+ "model.layers.14.self_attn.q_proj.weight": "model.safetensors",
73
+ "model.layers.14.self_attn.v_proj.weight": "model.safetensors",
74
+ "model.layers.15.conv.conv.weight": "model.safetensors",
75
+ "model.layers.15.conv.in_proj.weight": "model.safetensors",
76
+ "model.layers.15.conv.out_proj.weight": "model.safetensors",
77
+ "model.layers.15.feed_forward.w1.weight": "model.safetensors",
78
+ "model.layers.15.feed_forward.w2.weight": "model.safetensors",
79
+ "model.layers.15.feed_forward.w3.weight": "model.safetensors",
80
+ "model.layers.15.ffn_norm.weight": "model.safetensors",
81
+ "model.layers.15.operator_norm.weight": "model.safetensors",
82
+ "model.layers.2.feed_forward.w1.weight": "model.safetensors",
83
+ "model.layers.2.feed_forward.w2.weight": "model.safetensors",
84
+ "model.layers.2.feed_forward.w3.weight": "model.safetensors",
85
+ "model.layers.2.ffn_norm.weight": "model.safetensors",
86
+ "model.layers.2.operator_norm.weight": "model.safetensors",
87
+ "model.layers.2.self_attn.k_layernorm.weight": "model.safetensors",
88
+ "model.layers.2.self_attn.k_proj.weight": "model.safetensors",
89
+ "model.layers.2.self_attn.out_proj.weight": "model.safetensors",
90
+ "model.layers.2.self_attn.q_layernorm.weight": "model.safetensors",
91
+ "model.layers.2.self_attn.q_proj.weight": "model.safetensors",
92
+ "model.layers.2.self_attn.v_proj.weight": "model.safetensors",
93
+ "model.layers.3.conv.conv.weight": "model.safetensors",
94
+ "model.layers.3.conv.in_proj.weight": "model.safetensors",
95
+ "model.layers.3.conv.out_proj.weight": "model.safetensors",
96
+ "model.layers.3.feed_forward.w1.weight": "model.safetensors",
97
+ "model.layers.3.feed_forward.w2.weight": "model.safetensors",
98
+ "model.layers.3.feed_forward.w3.weight": "model.safetensors",
99
+ "model.layers.3.ffn_norm.weight": "model.safetensors",
100
+ "model.layers.3.operator_norm.weight": "model.safetensors",
101
+ "model.layers.4.conv.conv.weight": "model.safetensors",
102
+ "model.layers.4.conv.in_proj.weight": "model.safetensors",
103
+ "model.layers.4.conv.out_proj.weight": "model.safetensors",
104
+ "model.layers.4.feed_forward.w1.weight": "model.safetensors",
105
+ "model.layers.4.feed_forward.w2.weight": "model.safetensors",
106
+ "model.layers.4.feed_forward.w3.weight": "model.safetensors",
107
+ "model.layers.4.ffn_norm.weight": "model.safetensors",
108
+ "model.layers.4.operator_norm.weight": "model.safetensors",
109
+ "model.layers.5.feed_forward.w1.weight": "model.safetensors",
110
+ "model.layers.5.feed_forward.w2.weight": "model.safetensors",
111
+ "model.layers.5.feed_forward.w3.weight": "model.safetensors",
112
+ "model.layers.5.ffn_norm.weight": "model.safetensors",
113
+ "model.layers.5.operator_norm.weight": "model.safetensors",
114
+ "model.layers.5.self_attn.k_layernorm.weight": "model.safetensors",
115
+ "model.layers.5.self_attn.k_proj.weight": "model.safetensors",
116
+ "model.layers.5.self_attn.out_proj.weight": "model.safetensors",
117
+ "model.layers.5.self_attn.q_layernorm.weight": "model.safetensors",
118
+ "model.layers.5.self_attn.q_proj.weight": "model.safetensors",
119
+ "model.layers.5.self_attn.v_proj.weight": "model.safetensors",
120
+ "model.layers.6.conv.conv.weight": "model.safetensors",
121
+ "model.layers.6.conv.in_proj.weight": "model.safetensors",
122
+ "model.layers.6.conv.out_proj.weight": "model.safetensors",
123
+ "model.layers.6.feed_forward.w1.weight": "model.safetensors",
124
+ "model.layers.6.feed_forward.w2.weight": "model.safetensors",
125
+ "model.layers.6.feed_forward.w3.weight": "model.safetensors",
126
+ "model.layers.6.ffn_norm.weight": "model.safetensors",
127
+ "model.layers.6.operator_norm.weight": "model.safetensors",
128
+ "model.layers.7.conv.conv.weight": "model.safetensors",
129
+ "model.layers.7.conv.in_proj.weight": "model.safetensors",
130
+ "model.layers.7.conv.out_proj.weight": "model.safetensors",
131
+ "model.layers.7.feed_forward.w1.weight": "model.safetensors",
132
+ "model.layers.7.feed_forward.w2.weight": "model.safetensors",
133
+ "model.layers.7.feed_forward.w3.weight": "model.safetensors",
134
+ "model.layers.7.ffn_norm.weight": "model.safetensors",
135
+ "model.layers.7.operator_norm.weight": "model.safetensors",
136
+ "model.layers.8.feed_forward.w1.weight": "model.safetensors",
137
+ "model.layers.8.feed_forward.w2.weight": "model.safetensors",
138
+ "model.layers.8.feed_forward.w3.weight": "model.safetensors",
139
+ "model.layers.8.ffn_norm.weight": "model.safetensors",
140
+ "model.layers.8.operator_norm.weight": "model.safetensors",
141
+ "model.layers.8.self_attn.k_layernorm.weight": "model.safetensors",
142
+ "model.layers.8.self_attn.k_proj.weight": "model.safetensors",
143
+ "model.layers.8.self_attn.out_proj.weight": "model.safetensors",
144
+ "model.layers.8.self_attn.q_layernorm.weight": "model.safetensors",
145
+ "model.layers.8.self_attn.q_proj.weight": "model.safetensors",
146
+ "model.layers.8.self_attn.v_proj.weight": "model.safetensors",
147
+ "model.layers.9.conv.conv.weight": "model.safetensors",
148
+ "model.layers.9.conv.in_proj.weight": "model.safetensors",
149
+ "model.layers.9.conv.out_proj.weight": "model.safetensors",
150
+ "model.layers.9.feed_forward.w1.weight": "model.safetensors",
151
+ "model.layers.9.feed_forward.w2.weight": "model.safetensors",
152
+ "model.layers.9.feed_forward.w3.weight": "model.safetensors",
153
+ "model.layers.9.ffn_norm.weight": "model.safetensors",
154
+ "model.layers.9.operator_norm.weight": "model.safetensors"
155
+ }
156
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "bos_token": "<|startoftext|>",
4
+ "clean_up_tokenization_spaces": false,
5
+ "eos_token": "<|im_end|>",
6
+ "is_local": true,
7
+ "legacy": false,
8
+ "model_input_names": [
9
+ "input_ids",
10
+ "attention_mask"
11
+ ],
12
+ "model_max_length": 1000000000000000019884624838656,
13
+ "pad_token": "<|pad|>",
14
+ "sp_model_kwargs": {},
15
+ "spaces_between_special_tokens": false,
16
+ "tokenizer_class": "TokenizersBackend",
17
+ "use_default_system_prompt": false,
18
+ "use_fast": true
19
+ }