Hugging Face's logo

async0x42
/
GLM-4.7-Flash-4bit

Text Generation
MLX
Safetensors
English
Chinese
glm4_moe_lite
unsloth
conversational
4-bit precision
Model card Files
xet
 Community
async0x42 commited on
Commit
aaa1cd7
·
verified ·
1 Parent(s): 78e56fa

Add files using upload-large-folder tool

Browse files
Files changed (12) hide show
  1. .gitattributes +1 -0
  2. README.md +40 -0
  3. chat_template.jinja +86 -0
  4. config.json +115 -0
  5. generation_config.json +13 -0
  6. model-00001-of-00004.safetensors +3 -0
  7. model-00002-of-00004.safetensors +3 -0
  8. model-00003-of-00004.safetensors +3 -0
  9. model-00004-of-00004.safetensors +3 -0
  10. model.safetensors.index.json +0 -0
  11. tokenizer.json +3 -0
  12. tokenizer_config.json +35 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - unsloth
4
+ - mlx
5
+ base_model: unsloth/GLM-4.7-Flash
6
+ language:
7
+ - en
8
+ - zh
9
+ library_name: mlx
10
+ license: mit
11
+ pipeline_tag: text-generation
12
+ ---
13
+
14
+ # async0x42/GLM-4.7-Flash-4bit
15
+
16
+ This model [async0x42/GLM-4.7-Flash-4bit](https://huggingface.co/async0x42/GLM-4.7-Flash-4bit) was
17
+ converted to MLX format from [unsloth/GLM-4.7-Flash](https://huggingface.co/unsloth/GLM-4.7-Flash)
18
+ using mlx-lm version **0.30.7**.
19
+
20
+ ## Use with mlx
21
+
22
+ ```bash
23
+ pip install mlx-lm
24
+ ```
25
+
26
+ ```python
27
+ from mlx_lm import load, generate
28
+
29
+ model, tokenizer = load("async0x42/GLM-4.7-Flash-4bit")
30
+
31
+ prompt = "hello"
32
+
33
+ if tokenizer.chat_template is not None:
34
+ messages = [{"role": "user", "content": prompt}]
35
+ prompt = tokenizer.apply_chat_template(
36
+ messages, add_generation_prompt=True, return_dict=False,
37
+ )
38
+
39
+ response = generate(model, tokenizer, prompt=prompt, verbose=True)
40
+ ```
chat_template.jinja ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [gMASK]<sop>
2
+ {%- if tools -%}
3
+ <|system|>
4
+ # Tools
5
+
6
+ You may call one or more functions to assist with the user query.
7
+
8
+ You are provided with function signatures within <tools></tools> XML tags:
9
+ <tools>
10
+ {% for tool in tools %}
11
+ {{ tool | tojson(ensure_ascii=False) }}
12
+ {% endfor %}
13
+ </tools>
14
+
15
+ For each function call, output the function name and arguments within the following XML format:
16
+ <tool_call>{function-name}<arg_key>{arg-key-1}</arg_key><arg_value>{arg-value-1}</arg_value><arg_key>{arg-key-2}</arg_key><arg_value>{arg-value-2}</arg_value>...</tool_call>{%- endif -%}
17
+ {%- macro visible_text(content) -%}
18
+ {%- if content is string -%}
19
+ {{- content }}
20
+ {%- elif content is iterable and content is not mapping -%}
21
+ {%- for item in content -%}
22
+ {%- if item is mapping and item.type == 'text' -%}
23
+ {{- item.text }}
24
+ {%- elif item is string -%}
25
+ {{- item }}
26
+ {%- endif -%}
27
+ {%- endfor -%}
28
+ {%- else -%}
29
+ {{- content }}
30
+ {%- endif -%}
31
+ {%- endmacro -%}
32
+ {%- set ns = namespace(last_user_index=-1) %}
33
+ {%- for m in messages %}
34
+ {%- if m.role == 'user' %}
35
+ {% set ns.last_user_index = loop.index0 -%}
36
+ {%- endif %}
37
+ {%- endfor %}
38
+ {% for m in messages %}
39
+ {%- if m.role == 'user' -%}<|user|>{{ visible_text(m.content) }}
40
+ {%- elif m.role == 'assistant' -%}
41
+ <|assistant|>
42
+ {%- set reasoning_content = '' %}
43
+ {%- set content = visible_text(m.content) %}
44
+ {%- if m.reasoning_content is string %}
45
+ {%- set reasoning_content = m.reasoning_content %}
46
+ {%- else %}
47
+ {%- if '</think>' in content %}
48
+ {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
49
+ {%- set content = content.split('</think>')[-1].lstrip('\n') %}
50
+ {%- endif %}
51
+ {%- endif %}
52
+ {%- if ((clear_thinking is defined and not clear_thinking) or loop.index0 > ns.last_user_index) and reasoning_content -%}
53
+ {{ '<think>' + reasoning_content.strip() + '</think>'}}
54
+ {%- else -%}
55
+ {{ '</think>' }}
56
+ {%- endif -%}
57
+ {%- if content.strip() -%}
58
+ {{ content.strip() }}
59
+ {%- endif -%}
60
+ {% if m.tool_calls %}
61
+ {% for tc in m.tool_calls %}
62
+ {%- if tc.function %}
63
+ {%- set tc = tc.function %}
64
+ {%- endif %}
65
+ {{- '<tool_call>' + tc.name -}}
66
+ {% set _args = tc.arguments %}{% for k, v in _args.items() %}<arg_key>{{ k }}</arg_key><arg_value>{{ v | tojson(ensure_ascii=False) if v is not string else v }}</arg_value>{% endfor %}</tool_call>{% endfor %}
67
+ {% endif %}
68
+ {%- elif m.role == 'tool' -%}
69
+ {%- if m.content is string -%}
70
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
71
+ {{- '<|observation|>' }}
72
+ {%- endif %}
73
+ {{- '<tool_response>' }}
74
+ {{- m.content }}
75
+ {{- '</tool_response>' }}
76
+ {%- else -%}
77
+ <|observation|>{% for tr in m.content %}
78
+ <tool_response>{{ tr.output if tr.output is defined else tr }}</tool_response>{% endfor -%}
79
+ {% endif -%}
80
+ {%- elif m.role == 'system' -%}
81
+ <|system|>{{ visible_text(m.content) }}
82
+ {%- endif -%}
83
+ {%- endfor -%}
84
+ {%- if add_generation_prompt -%}
85
+ <|assistant|>{{- '</think>' if (enable_thinking is defined and not enable_thinking) else '<think>' -}}
86
+ {%- endif -%}
config.json ADDED
@@ -0,0 +1,115 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Glm4MoeLiteForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "eos_token_id": [
8
+ 154820,
9
+ 154827,
10
+ 154829
11
+ ],
12
+ "first_k_dense_replace": 1,
13
+ "head_dim": 64,
14
+ "hidden_act": "silu",
15
+ "hidden_size": 2048,
16
+ "initializer_range": 0.02,
17
+ "intermediate_size": 10240,
18
+ "kv_lora_rank": 512,
19
+ "max_position_embeddings": 202752,
20
+ "mlp_layer_types": [
21
+ "dense",
22
+ "sparse",
23
+ "sparse",
24
+ "sparse",
25
+ "sparse",
26
+ "sparse",
27
+ "sparse",
28
+ "sparse",
29
+ "sparse",
30
+ "sparse",
31
+ "sparse",
32
+ "sparse",
33
+ "sparse",
34
+ "sparse",
35
+ "sparse",
36
+ "sparse",
37
+ "sparse",
38
+ "sparse",
39
+ "sparse",
40
+ "sparse",
41
+ "sparse",
42
+ "sparse",
43
+ "sparse",
44
+ "sparse",
45
+ "sparse",
46
+ "sparse",
47
+ "sparse",
48
+ "sparse",
49
+ "sparse",
50
+ "sparse",
51
+ "sparse",
52
+ "sparse",
53
+ "sparse",
54
+ "sparse",
55
+ "sparse",
56
+ "sparse",
57
+ "sparse",
58
+ "sparse",
59
+ "sparse",
60
+ "sparse",
61
+ "sparse",
62
+ "sparse",
63
+ "sparse",
64
+ "sparse",
65
+ "sparse",
66
+ "sparse",
67
+ "sparse"
68
+ ],
69
+ "model_type": "glm4_moe_lite",
70
+ "moe_intermediate_size": 1536,
71
+ "n_group": 1,
72
+ "n_routed_experts": 64,
73
+ "n_shared_experts": 1,
74
+ "norm_topk_prob": true,
75
+ "num_attention_heads": 20,
76
+ "num_experts_per_tok": 4,
77
+ "num_hidden_layers": 47,
78
+ "num_key_value_heads": 20,
79
+ "num_nextn_predict_layers": 1,
80
+ "pad_token_id": 154821,
81
+ "partial_rotary_factor": 1.0,
82
+ "pretraining_tp": 1,
83
+ "q_lora_rank": 768,
84
+ "qk_head_dim": 256,
85
+ "qk_nope_head_dim": 192,
86
+ "qk_rope_head_dim": 64,
87
+ "quantization": {
88
+ "group_size": 64,
89
+ "bits": 4,
90
+ "mode": "affine"
91
+ },
92
+ "quantization_config": {
93
+ "group_size": 64,
94
+ "bits": 4,
95
+ "mode": "affine"
96
+ },
97
+ "rms_norm_eps": 1e-05,
98
+ "rope_interleave": true,
99
+ "rope_parameters": {
100
+ "partial_rotary_factor": 1.0,
101
+ "rope_theta": 1000000,
102
+ "rope_type": "default"
103
+ },
104
+ "routed_scaling_factor": 1.8,
105
+ "scoring_func": "sigmoid",
106
+ "tie_word_embeddings": false,
107
+ "topk_group": 1,
108
+ "topk_method": "noaux_tc",
109
+ "torch_dtype": "bfloat16",
110
+ "transformers_version": "5.2.0.dev0",
111
+ "unsloth_fixed": true,
112
+ "use_cache": true,
113
+ "v_head_dim": 256,
114
+ "vocab_size": 154880
115
+ }
generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "eos_token_id": [
4
+ 154820,
5
+ 154827,
6
+ 154829
7
+ ],
8
+ "pad_token_id": 154820,
9
+ "temperature": 1.0,
10
+ "top_p" : 0.95,
11
+ "do_sample": true,
12
+ "transformers_version": "5.0.0.dev0"
13
+ }
model-00001-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:540741ab088c32f7338636c82c601399c0db6c11c965ac2c2865e9400d32dd8b
3
+ size 5357453035
model-00002-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28749aa10a019ffa2a68e12a94c3883e5102942b310937667bb359221a4d4a8d
3
+ size 5363462379
model-00003-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c6896cb4c209344aac52eb9784d34145bf752aa99653d480d211f8261d5449b
3
+ size 5363462333
model-00004-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48361cae86bb6d78ac246a1a23ec92e863eef276e9bc285ec6a7cbb5f8bc0f07
3
+ size 768064072
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19e773648cb4e65de8660ea6365e10acca112d42a854923df93db4a6f333a82d
3
+ size 20217442
tokenizer_config.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "bos_token": null,
4
+ "clean_up_tokenization_spaces": false,
5
+ "do_lower_case": false,
6
+ "eos_token": "<|endoftext|>",
7
+ "extra_special_tokens": [
8
+ "<|endoftext|>",
9
+ "[MASK]",
10
+ "[gMASK]",
11
+ "[sMASK]",
12
+ "<sop>",
13
+ "<eop>",
14
+ "<|system|>",
15
+ "<|user|>",
16
+ "<|assistant|>",
17
+ "<|observation|>",
18
+ "<|begin_of_image|>",
19
+ "<|end_of_image|>",
20
+ "<|begin_of_video|>",
21
+ "<|end_of_video|>",
22
+ "<|begin_of_audio|>",
23
+ "<|end_of_audio|>",
24
+ "<|begin_of_transcription|>",
25
+ "<|end_of_transcription|>"
26
+ ],
27
+ "is_local": true,
28
+ "model_max_length": 202752,
29
+ "pad_token": "[MASK]",
30
+ "padding_side": "left",
31
+ "remove_space": false,
32
+ "tokenizer_class": "TokenizersBackend",
33
+ "tool_parser_type": "glm47",
34
+ "unk_token": null
35
+ }