Hugging Face's logo

samadpls
/
torchlight-phi4-mini-lora

Text Generation
PEFT
Safetensors
PyTorch
MLX
English
code
debugging
lora
phi3
Model card Files
xet
 Community
samadpls commited on
Commit
f4e45ca
·
verified ·
1 Parent(s): 8adca30

Add the LoRA adapter in PEFT and mlx-lm formats

Browse files
Files changed (4) hide show
  1. adapter_config.json +77 -0
  2. adapter_model.safetensors +3 -0
  3. mlx/adapter_config.json +48 -0
  4. mlx/adapters.safetensors +3 -0
adapter_config.json ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "peft_type": "LORA",
3
+ "task_type": "CAUSAL_LM",
4
+ "base_model_name_or_path": "microsoft/Phi-4-mini-instruct",
5
+ "r": 16,
6
+ "lora_alpha": 32.0,
7
+ "lora_dropout": 0.0,
8
+ "bias": "none",
9
+ "fan_in_fan_out": false,
10
+ "inference_mode": true,
11
+ "target_modules": [
12
+ "model.layers.16.mlp.down_proj",
13
+ "model.layers.16.mlp.gate_up_proj",
14
+ "model.layers.16.self_attn.o_proj",
15
+ "model.layers.16.self_attn.qkv_proj",
16
+ "model.layers.17.mlp.down_proj",
17
+ "model.layers.17.mlp.gate_up_proj",
18
+ "model.layers.17.self_attn.o_proj",
19
+ "model.layers.17.self_attn.qkv_proj",
20
+ "model.layers.18.mlp.down_proj",
21
+ "model.layers.18.mlp.gate_up_proj",
22
+ "model.layers.18.self_attn.o_proj",
23
+ "model.layers.18.self_attn.qkv_proj",
24
+ "model.layers.19.mlp.down_proj",
25
+ "model.layers.19.mlp.gate_up_proj",
26
+ "model.layers.19.self_attn.o_proj",
27
+ "model.layers.19.self_attn.qkv_proj",
28
+ "model.layers.20.mlp.down_proj",
29
+ "model.layers.20.mlp.gate_up_proj",
30
+ "model.layers.20.self_attn.o_proj",
31
+ "model.layers.20.self_attn.qkv_proj",
32
+ "model.layers.21.mlp.down_proj",
33
+ "model.layers.21.mlp.gate_up_proj",
34
+ "model.layers.21.self_attn.o_proj",
35
+ "model.layers.21.self_attn.qkv_proj",
36
+ "model.layers.22.mlp.down_proj",
37
+ "model.layers.22.mlp.gate_up_proj",
38
+ "model.layers.22.self_attn.o_proj",
39
+ "model.layers.22.self_attn.qkv_proj",
40
+ "model.layers.23.mlp.down_proj",
41
+ "model.layers.23.mlp.gate_up_proj",
42
+ "model.layers.23.self_attn.o_proj",
43
+ "model.layers.23.self_attn.qkv_proj",
44
+ "model.layers.24.mlp.down_proj",
45
+ "model.layers.24.mlp.gate_up_proj",
46
+ "model.layers.24.self_attn.o_proj",
47
+ "model.layers.24.self_attn.qkv_proj",
48
+ "model.layers.25.mlp.down_proj",
49
+ "model.layers.25.mlp.gate_up_proj",
50
+ "model.layers.25.self_attn.o_proj",
51
+ "model.layers.25.self_attn.qkv_proj",
52
+ "model.layers.26.mlp.down_proj",
53
+ "model.layers.26.mlp.gate_up_proj",
54
+ "model.layers.26.self_attn.o_proj",
55
+ "model.layers.26.self_attn.qkv_proj",
56
+ "model.layers.27.mlp.down_proj",
57
+ "model.layers.27.mlp.gate_up_proj",
58
+ "model.layers.27.self_attn.o_proj",
59
+ "model.layers.27.self_attn.qkv_proj",
60
+ "model.layers.28.mlp.down_proj",
61
+ "model.layers.28.mlp.gate_up_proj",
62
+ "model.layers.28.self_attn.o_proj",
63
+ "model.layers.28.self_attn.qkv_proj",
64
+ "model.layers.29.mlp.down_proj",
65
+ "model.layers.29.mlp.gate_up_proj",
66
+ "model.layers.29.self_attn.o_proj",
67
+ "model.layers.29.self_attn.qkv_proj",
68
+ "model.layers.30.mlp.down_proj",
69
+ "model.layers.30.mlp.gate_up_proj",
70
+ "model.layers.30.self_attn.o_proj",
71
+ "model.layers.30.self_attn.qkv_proj",
72
+ "model.layers.31.mlp.down_proj",
73
+ "model.layers.31.mlp.gate_up_proj",
74
+ "model.layers.31.self_attn.o_proj",
75
+ "model.layers.31.self_attn.qkv_proj"
76
+ ]
77
+ }
adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5dfb61ff19c292717964769fcb688eab05cbdca031f1bdc8c62b916f6d30dbb
3
+ size 46154552
mlx/adapter_config.json ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "adapter_path": "data/train/phi4-mini/adapters",
3
+ "batch_size": 2,
4
+ "clear_cache_threshold": 0,
5
+ "config": "data/train/phi4-mini/lora.yaml",
6
+ "data": "data/train/phi4-mini/data",
7
+ "fine_tune_type": "lora",
8
+ "grad_accumulation_steps": 4,
9
+ "grad_checkpoint": true,
10
+ "iters": 3459,
11
+ "learning_rate": 0.0002,
12
+ "lora_parameters": {
13
+ "rank": 16,
14
+ "scale": 2.0,
15
+ "dropout": 0.0
16
+ },
17
+ "lr_schedule": {
18
+ "name": "cosine_decay",
19
+ "warmup": 43,
20
+ "arguments": [
21
+ 0.0002,
22
+ 821,
23
+ 0.0
24
+ ]
25
+ },
26
+ "mask_prompt": true,
27
+ "max_seq_length": 4096,
28
+ "model": "microsoft/Phi-4-mini-instruct",
29
+ "num_layers": 16,
30
+ "optimizer": "adamw",
31
+ "optimizer_config": {
32
+ "adamw": {
33
+ "weight_decay": 0.01
34
+ }
35
+ },
36
+ "project_name": null,
37
+ "report_to": null,
38
+ "resume_adapter_file": null,
39
+ "save_every": 1000,
40
+ "seed": 0,
41
+ "steps_per_eval": 1000,
42
+ "steps_per_report": 10,
43
+ "test": false,
44
+ "test_batches": 500,
45
+ "train": true,
46
+ "trust_remote_code": false,
47
+ "val_batches": -1
48
+ }
mlx/adapters.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e39525e0a72f712cce4a6402522da5f3265af95d925bb5b1e6f6644a9352ee7d
3
+ size 46151500