Hugging Face's logo

software-mansion
/
Qwen3-ExecuTorch

Text Generation
ExecuTorch
conversational
Model card Files
xet
 Community
1
msluszniak commited on
Commit
fa72381
·
verified ·
1 Parent(s): e81de1f

Add MLX int4 variants for 0.6B, 1.7B and 4B

Browse files
Files changed (7) hide show
  1. .gitattributes +3 -0
  2. 0_6b/mlx/config.json +61 -0
  3. 0_6b/mlx/qwen_3_0_6b_mlx_int4.pte +3 -0
  4. 1_7b/mlx/config.json +61 -0
  5. 1_7b/mlx/qwen_3_1_7b_mlx_int4.pte +3 -0
  6. 4b/mlx/config.json +61 -0
  7. 4b/mlx/qwen_3_4b_mlx_int4.pte +3 -0
.gitattributes CHANGED
@@ -46,3 +46,6 @@ qwen-3-4B/quantized/qwen3_4b_8da4w.pte filter=lfs diff=lfs merge=lfs -text
46
  1_7b/xnnpack/qwen_3_1_7b_xnnpack_8da4w.pte filter=lfs diff=lfs merge=lfs -text
47
  4b/xnnpack/qwen_3_4b_xnnpack_bf16.pte filter=lfs diff=lfs merge=lfs -text
48
  4b/xnnpack/qwen_3_4b_xnnpack_8da4w.pte filter=lfs diff=lfs merge=lfs -text
 
 
 
 
46
  1_7b/xnnpack/qwen_3_1_7b_xnnpack_8da4w.pte filter=lfs diff=lfs merge=lfs -text
47
  4b/xnnpack/qwen_3_4b_xnnpack_bf16.pte filter=lfs diff=lfs merge=lfs -text
48
  4b/xnnpack/qwen_3_4b_xnnpack_8da4w.pte filter=lfs diff=lfs merge=lfs -text
49
+ 0_6b/mlx/qwen_3_0_6b_mlx_int4.pte filter=lfs diff=lfs merge=lfs -text
50
+ 1_7b/mlx/qwen_3_1_7b_mlx_int4.pte filter=lfs diff=lfs merge=lfs -text
51
+ 4b/mlx/qwen_3_4b_mlx_int4.pte filter=lfs diff=lfs merge=lfs -text
0_6b/mlx/config.json ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
3
+ "model": "qwen_3",
4
+ "family": "qwen",
5
+ "capabilities": [
6
+ "text-generation"
7
+ ],
8
+ "backend": "mlx",
9
+ "license": "apache-2.0",
10
+ "variants": [
11
+ {
12
+ "file": "qwen_3_0_6b_mlx_int4.pte",
13
+ "precision": "4w",
14
+ "quantized": true,
15
+ "default": true,
16
+ "methods": {
17
+ "enable_dynamic_shape": true,
18
+ "forward": {
19
+ "inputs": [
20
+ {
21
+ "shape": [
22
+ 1,
23
+ {
24
+ "min": 1,
25
+ "max": 2047
26
+ }
27
+ ],
28
+ "dtype": "int64"
29
+ },
30
+ {
31
+ "shape": [
32
+ 1
33
+ ],
34
+ "dtype": "int64"
35
+ }
36
+ ],
37
+ "outputs": [
38
+ {
39
+ "shape": [
40
+ 1,
41
+ 151936
42
+ ],
43
+ "dtype": "bfloat16"
44
+ }
45
+ ]
46
+ },
47
+ "get_bos_id": 151644,
48
+ "get_eos_ids": [
49
+ 151645
50
+ ],
51
+ "get_max_context_len": 2048,
52
+ "get_max_seq_len": 2048,
53
+ "get_n_layers": 28,
54
+ "get_vocab_size": 151936,
55
+ "use_kv_cache": true,
56
+ "use_sdpa_with_kv_cache": true
57
+ }
58
+ }
59
+ ],
60
+ "size": "0_6b"
61
+ }
0_6b/mlx/qwen_3_0_6b_mlx_int4.pte ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1aa68502ee4557349009817fb676c139d842352efc52d77f1b65c2347e71dd7
3
+ size 628164608
1_7b/mlx/config.json ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
3
+ "model": "qwen_3",
4
+ "family": "qwen",
5
+ "capabilities": [
6
+ "text-generation"
7
+ ],
8
+ "backend": "mlx",
9
+ "license": "apache-2.0",
10
+ "variants": [
11
+ {
12
+ "file": "qwen_3_1_7b_mlx_int4.pte",
13
+ "precision": "4w",
14
+ "quantized": true,
15
+ "default": true,
16
+ "methods": {
17
+ "enable_dynamic_shape": true,
18
+ "forward": {
19
+ "inputs": [
20
+ {
21
+ "shape": [
22
+ 1,
23
+ {
24
+ "min": 1,
25
+ "max": 2047
26
+ }
27
+ ],
28
+ "dtype": "int64"
29
+ },
30
+ {
31
+ "shape": [
32
+ 1
33
+ ],
34
+ "dtype": "int64"
35
+ }
36
+ ],
37
+ "outputs": [
38
+ {
39
+ "shape": [
40
+ 1,
41
+ 151936
42
+ ],
43
+ "dtype": "bfloat16"
44
+ }
45
+ ]
46
+ },
47
+ "get_bos_id": 151644,
48
+ "get_eos_ids": [
49
+ 151645
50
+ ],
51
+ "get_max_context_len": 2048,
52
+ "get_max_seq_len": 2048,
53
+ "get_n_layers": 28,
54
+ "get_vocab_size": 151936,
55
+ "use_kv_cache": true,
56
+ "use_sdpa_with_kv_cache": true
57
+ }
58
+ }
59
+ ],
60
+ "size": "1_7b"
61
+ }
1_7b/mlx/qwen_3_1_7b_mlx_int4.pte ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac12a5190f510c20a8fabe0e5370766e434f3db68360c1ce49f5936084f777d2
3
+ size 1536819200
4b/mlx/config.json ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
3
+ "model": "qwen_3",
4
+ "family": "qwen",
5
+ "capabilities": [
6
+ "text-generation"
7
+ ],
8
+ "backend": "mlx",
9
+ "license": "apache-2.0",
10
+ "variants": [
11
+ {
12
+ "file": "qwen_3_4b_mlx_int4.pte",
13
+ "precision": "4w",
14
+ "quantized": true,
15
+ "default": true,
16
+ "methods": {
17
+ "enable_dynamic_shape": true,
18
+ "forward": {
19
+ "inputs": [
20
+ {
21
+ "shape": [
22
+ 1,
23
+ {
24
+ "min": 1,
25
+ "max": 2047
26
+ }
27
+ ],
28
+ "dtype": "int64"
29
+ },
30
+ {
31
+ "shape": [
32
+ 1
33
+ ],
34
+ "dtype": "int64"
35
+ }
36
+ ],
37
+ "outputs": [
38
+ {
39
+ "shape": [
40
+ 1,
41
+ 151936
42
+ ],
43
+ "dtype": "bfloat16"
44
+ }
45
+ ]
46
+ },
47
+ "get_bos_id": 151644,
48
+ "get_eos_ids": [
49
+ 151645
50
+ ],
51
+ "get_max_context_len": 2048,
52
+ "get_max_seq_len": 2048,
53
+ "get_n_layers": 36,
54
+ "get_vocab_size": 151936,
55
+ "use_kv_cache": true,
56
+ "use_sdpa_with_kv_cache": true
57
+ }
58
+ }
59
+ ],
60
+ "size": "4b"
61
+ }
4b/mlx/qwen_3_4b_mlx_int4.pte ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:296dc9be0b5092405e10517b40808212bef9b88f7273d002d3a9c3bfcf488e7e
3
+ size 2915458688