gss1147 commited on
Commit
13ca4f5
·
verified ·
1 Parent(s): ac3b4f6

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +205 -51
README.md CHANGED
@@ -1,59 +1,213 @@
1
  ---
2
- base_model:
3
- - LucidityAI/Astral-4B-Coder
4
- - openfree/Darwin-Qwen3-4B
5
- - Qwen/Qwen3-4B
6
  library_name: transformers
 
 
 
 
7
  tags:
8
- - mergekit
9
- - merge
 
 
 
 
 
 
 
10
  datasets:
11
- - LucidityAI/Astral-Post-Training-Dataset
 
12
  ---
13
- # Within Us Ai Hybrid Pick merge
14
 
15
- This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
16
 
17
  ## Merge Details
18
- ### Merge Method
19
-
20
- This model was merged using the [SLERP](https://en.wikipedia.org/wiki/Slerp) merge method.
21
-
22
- ### Models Merged
23
-
24
- The following models were included in the merge:
25
- * [openfree-Darwin-Qwen3-4B](https://huggingface.co/openfree/Darwin-Qwen3-4B/tree/main)
26
- * [Luciditydd-AI-Astral-4B-Coder](https://huggingface.co/LucidityAI/Astral-4B-Coder/tree/main)
27
-
28
-
29
- ### Configuration
30
-
31
- The following YAML configuration was used to produce this model:
32
-
33
- ```yaml
34
- base_model:Qwen/Qwen3-4B-Instruct-2507
35
- dtype: float16
36
- merge_method: slerp
37
- parameters:
38
- t:
39
- - filter: embed_tokens
40
- value: 0.0
41
- - filter: self_attn
42
- value: 0.5
43
- - filter: mlp
44
- value: 0.5
45
- - filter: lm_head
46
- value: 1.0
47
- - value: 0.5
48
- slices:
49
- - sources:
50
- - layer_range:
51
- - 0
52
- - 36
53
- model: Lucidity-AI-Astral-4B-Coder
54
- - layer_range:
55
- - 0
56
- - 36
57
- model: openfree-Darwin-Qwen3-4B
58
-
59
- ```
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ license: other
 
 
 
3
  library_name: transformers
4
+ base_model:
5
+ - LucidityAI/Astral-4B-Coder
6
+ - openfree/Darwin-Qwen3-4B
7
+ - Qwen/Qwen3-4B
8
  tags:
9
+ - qwen3
10
+ - mergekit
11
+ - merge
12
+ - text-generation-inference
13
+ - code
14
+ - coder
15
+ - withinusai
16
+ language:
17
+ - en
18
  datasets:
19
+ - LucidityAI/Astral-Post-Training-Dataset
20
+ pipeline_tag: text-generation
21
  ---
 
22
 
23
+ # Darwin-Astral-4B-Coder
24
+
25
+ **Darwin-Astral-4B-Coder** is a merged 4B-class coding model release from **WithIn Us AI**, designed for code generation, instruction-following, and practical developer-assistant workflows.
26
+
27
+ This repository is distributed as a standard **Transformers** checkpoint in **Safetensors** format and is positioned as a merge-based model that blends Darwin-style and Astral-style coding traits within a Qwen3-family 4B backbone.
28
+
29
+ ## Model Summary
30
+
31
+ This model is intended for:
32
+
33
+ - code generation
34
+ - code explanation
35
+ - debugging assistance
36
+ - implementation planning
37
+ - instruction-following
38
+ - developer assistant workflows
39
+ - local or hosted coding inference
40
+
41
+ As a 4B-class model, it aims to balance stronger coding capability than very small models with a lighter deployment footprint than larger coder checkpoints.
42
+
43
+ ## Base Model Lineage
44
+
45
+ The current repository metadata lists the following upstream model references:
46
+
47
+ - `LucidityAI/Astral-4B-Coder`
48
+ - `openfree/Darwin-Qwen3-4B`
49
+ - `Qwen/Qwen3-4B`
50
+
51
+ The visible merge configuration in the README also shows:
52
+
53
+ - `Qwen/Qwen3-4B-Instruct-2507` as the base model in the YAML block
54
+ - `Lucidity-AI-Astral-4B-Coder` as a merge source
55
+ - `openfree-Darwin-Qwen3-4B` as a merge source
56
+
57
+ These names are preserved here as shown on the repository page.
58
 
59
  ## Merge Details
60
+
61
+ According to the current README:
62
+
63
+ - this model is a merge of pre-trained language models
64
+ - it was created using **mergekit**
65
+ - the **SLERP** merge method was used
66
+
67
+ The repository also includes a visible `mergekit_config.yml`, which supports the merge-based packaging of the release.
68
+
69
+ ## Dataset Lineage
70
+
71
+ The repository page currently shows the following dataset association:
72
+
73
+ - `LucidityAI/Astral-Post-Training-Dataset`
74
+
75
+ This suggests coding or post-training lineage connected to the Astral family used in the merge.
76
+
77
+ ## Intended Use
78
+
79
+ Recommended use cases include:
80
+
81
+ - coding assistant experiments
82
+ - generating utility functions and scripts
83
+ - explaining code and technical concepts
84
+ - debugging support
85
+ - step-by-step implementation planning
86
+ - local developer tools
87
+ - hosted text-generation workflows for software tasks
88
+
89
+ ## Suggested Use Cases
90
+
91
+ This model can be useful for:
92
+
93
+ - drafting Python, JavaScript, or general-purpose code
94
+ - proposing refactors
95
+ - generating boilerplate
96
+ - answering developer questions
97
+ - comparing implementation approaches
98
+ - producing structured technical responses
99
+
100
+ ## Out-of-Scope Use
101
+
102
+ This model should not be relied on for:
103
+
104
+ - legal advice
105
+ - medical advice
106
+ - financial advice
107
+ - safety-critical automation
108
+ - autonomous production engineering without review
109
+ - security-critical code without expert validation
110
+
111
+ All generated code should be reviewed, tested, and validated before real-world deployment.
112
+
113
+ ## Repository Contents
114
+
115
+ The repository currently includes standard Hugging Face model assets such as:
116
+
117
+ - `README.md`
118
+ - `.gitattributes`
119
+ - `added_tokens.json`
120
+ - `config.json`
121
+ - `mergekit_config.yml`
122
+ - `merges.txt`
123
+ - `model-00001-of-00002.safetensors`
124
+ - `model-00002-of-00002.safetensors`
125
+ - `model.safetensors.index.json`
126
+ - `special_tokens_map.json`
127
+ - `tokenizer.json`
128
+ - `tokenizer_config.json`
129
+
130
+ ## Prompting Guidance
131
+
132
+ This model will usually work best with prompts that are:
133
+
134
+ - direct
135
+ - scoped to a clear task
136
+ - explicit about the language or framework
137
+ - clear about whether code, explanation, or both are wanted
138
+ - structured when step-by-step reasoning is useful
139
+
140
+ ### Example prompt styles
141
+
142
+ **Code generation**
143
+ > Write a Python function that loads a JSON file, validates required keys, and returns cleaned records.
144
+
145
+ **Debugging**
146
+ > Explain why this code raises a KeyError and provide a safer corrected version.
147
+
148
+ **Implementation planning**
149
+ > Create a step-by-step plan for building a FastAPI service with authentication, logging, and tests.
150
+
151
+ **Refactoring**
152
+ > Refactor this function for readability and add basic error handling.
153
+
154
+ ## Strengths
155
+
156
+ This model may be especially useful for:
157
+
158
+ - blended coding workflows
159
+ - practical developer assistance
160
+ - moderate-size local inference
161
+ - structured software-task prompting
162
+ - merge-model experimentation
163
+ - compact coder deployments
164
+
165
+ ## Limitations
166
+
167
+ Like other merged 4B-class language models, this model may:
168
+
169
+ - hallucinate APIs or implementation details
170
+ - generate incomplete or incorrect code
171
+ - produce insecure patterns
172
+ - make reasoning mistakes on harder prompts
173
+ - require prompt iteration for best results
174
+ - need human validation before real-world use
175
+
176
+ ## Attribution
177
+
178
+ **WithIn Us AI** is the publisher of this merged model release.
179
+
180
+ Credit for upstream assets remains with their original creators. The repository metadata and README specifically reference:
181
+
182
+ - `LucidityAI/Astral-4B-Coder`
183
+ - `openfree/Darwin-Qwen3-4B`
184
+ - `Qwen/Qwen3-4B`
185
+ - `Qwen/Qwen3-4B-Instruct-2507`
186
+
187
+ and the dataset:
188
+
189
+ - `LucidityAI/Astral-Post-Training-Dataset`
190
+
191
+ ## License
192
+
193
+ This draft uses:
194
+
195
+ - `license: other`
196
+
197
+ If you maintain this repo, replace this with the exact license terms you want displayed and make sure they align with any upstream obligations from the referenced source models and datasets.
198
+
199
+ ## Acknowledgments
200
+
201
+ Thanks to:
202
+
203
+ - **WithIn Us AI**
204
+ - **LucidityAI**
205
+ - **openfree**
206
+ - **Qwen**
207
+ - the **mergekit** ecosystem
208
+ - the Hugging Face platform
209
+ - the broader open-source LLM community
210
+
211
+ ## Disclaimer
212
+
213
+ This model may produce inaccurate, insecure, biased, incomplete, or misleading outputs. All important generations, especially code and technical guidance, should be reviewed and tested before real-world use.