Feature Extraction
Transformers
Safetensors
ModularStarEncoder
custom_code
andreagurioli1995 commited on
Commit
0c0d4bb
·
verified ·
1 Parent(s): ae18bf0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +28 -188
README.md CHANGED
@@ -3,17 +3,22 @@ library_name: transformers
3
  datasets:
4
  - bigcode/the-stack-v2
5
  license: bigcode-openrail-m
 
 
6
  ---
7
 
8
- # Model Card for Model ID
9
 
10
  <!-- Provide a quick summary of what the model is/does. -->
11
 
 
 
 
12
 
 
13
 
14
- ## Model Details
15
-
16
-
17
 
18
  ### How to use
19
  ```python
@@ -53,189 +58,24 @@ You will get as an output three elements:
53
  - raw_hidden_states: raw representation from all the hidden states of the model, without pooling, normalization, and projection
54
  - attentions: attention scores from the encoder
55
 
56
- ### Model Description
 
57
 
58
  <!-- Provide a longer summary of what this model is. -->
59
-
60
- This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
61
-
62
- - **Developed by:** [More Information Needed]
63
- - **Funded by [optional]:** [More Information Needed]
64
- - **Shared by [optional]:** [More Information Needed]
65
- - **Model type:** [More Information Needed]
66
- - **Language(s) (NLP):** [More Information Needed]
67
- - **License:** [More Information Needed]
68
- - **Finetuned from model [optional]:** [More Information Needed]
69
-
70
- ### Model Sources [optional]
71
-
72
- <!-- Provide the basic links for the model. -->
73
-
74
- - **Repository:** [More Information Needed]
75
- - **Paper [optional]:** [More Information Needed]
76
- - **Demo [optional]:** [More Information Needed]
77
-
78
- ## Uses
79
-
80
- <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
81
-
82
- ### Direct Use
83
-
84
- <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
85
-
86
- [More Information Needed]
87
-
88
- ### Downstream Use [optional]
89
-
90
- <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
91
-
92
- [More Information Needed]
93
-
94
- ### Out-of-Scope Use
95
-
96
- <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
97
-
98
- [More Information Needed]
99
-
100
- ## Bias, Risks, and Limitations
101
-
102
- <!-- This section is meant to convey both technical and sociotechnical limitations. -->
103
-
104
- [More Information Needed]
105
-
106
- ### Recommendations
107
-
108
- <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
109
-
110
- Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
111
-
112
- ## How to Get Started with the Model
113
-
114
- Use the code below to get started with the model.
115
-
116
- [More Information Needed]
117
-
118
- ## Training Details
119
-
120
- ### Training Data
121
-
122
- <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
123
-
124
- [More Information Needed]
125
-
126
- ### Training Procedure
127
-
128
- <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
129
-
130
- #### Preprocessing [optional]
131
-
132
- [More Information Needed]
133
-
134
-
135
- #### Training Hyperparameters
136
-
137
- - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
138
-
139
- #### Speeds, Sizes, Times [optional]
140
-
141
- <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
142
-
143
- [More Information Needed]
144
-
145
- ## Evaluation
146
-
147
- <!-- This section describes the evaluation protocols and provides the results. -->
148
-
149
- ### Testing Data, Factors & Metrics
150
-
151
- #### Testing Data
152
-
153
- <!-- This should link to a Dataset Card if possible. -->
154
-
155
- [More Information Needed]
156
-
157
- #### Factors
158
-
159
- <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
160
-
161
- [More Information Needed]
162
-
163
- #### Metrics
164
-
165
- <!-- These are the evaluation metrics being used, ideally with a description of why. -->
166
-
167
- [More Information Needed]
168
-
169
- ### Results
170
-
171
- [More Information Needed]
172
-
173
- #### Summary
174
-
175
-
176
-
177
- ## Model Examination [optional]
178
-
179
- <!-- Relevant interpretability work for the model goes here -->
180
-
181
- [More Information Needed]
182
-
183
- ## Environmental Impact
184
-
185
- <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
186
-
187
- Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
188
-
189
- - **Hardware Type:** [More Information Needed]
190
- - **Hours used:** [More Information Needed]
191
- - **Cloud Provider:** [More Information Needed]
192
- - **Compute Region:** [More Information Needed]
193
- - **Carbon Emitted:** [More Information Needed]
194
-
195
- ## Technical Specifications [optional]
196
-
197
- ### Model Architecture and Objective
198
-
199
- [More Information Needed]
200
-
201
- ### Compute Infrastructure
202
-
203
- [More Information Needed]
204
-
205
- #### Hardware
206
-
207
- [More Information Needed]
208
-
209
- #### Software
210
-
211
- [More Information Needed]
212
-
213
- ## Citation [optional]
214
-
215
- <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
216
-
217
- **BibTeX:**
218
-
219
- [More Information Needed]
220
-
221
- **APA:**
222
-
223
- [More Information Needed]
224
-
225
- ## Glossary [optional]
226
-
227
- <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
228
-
229
- [More Information Needed]
230
-
231
- ## More Information [optional]
232
-
233
- [More Information Needed]
234
-
235
- ## Model Card Authors [optional]
236
-
237
- [More Information Needed]
238
-
239
- ## Model Card Contact
240
-
241
- [More Information Needed]
 
3
  datasets:
4
  - bigcode/the-stack-v2
5
  license: bigcode-openrail-m
6
+ base_model:
7
+ - andreagurioli1995/ModularStarEncoder
8
  ---
9
 
10
+ # ModularStarEncoder-1B Fine-Tuned model
11
 
12
  <!-- Provide a quick summary of what the model is/does. -->
13
 
14
+ ModularStarEncoder-finetuned is an encoder built on top of [ModularStarEncoder-1B Pre-trained](https://huggingface.co/andreagurioli1995/ModularStarEncoder) on [SynthCode2Code2NL](add link here).
15
+ ModularStarEncoder fine-tuned encoder for various retrieval tasks, enabling the end user to select the model size that meets their memory and computational constraints.
16
+ We built ModularStarEncoder on top of [StarCoder-2](https://huggingface.co/bigcode/starcoder2-15b), reducing its size from 15B to 1B parameters in bfloat16.
17
 
18
+ The model is finetuned with [CLIP objective](https://github.com/mlfoundations/open_clip/blob/main/src/open_clip/loss.py)
19
 
20
+ - **Paper:** [Link](arxiv.paper)
21
+ - **Languages:** English, Go, Ruby, Python, Java, C++, PHP, C, JavaScript
 
22
 
23
  ### How to use
24
  ```python
 
58
  - raw_hidden_states: raw representation from all the hidden states of the model, without pooling, normalization, and projection
59
  - attentions: attention scores from the encoder
60
 
61
+
62
+ ### Training
63
 
64
  <!-- Provide a longer summary of what this model is. -->
65
+ We fine-tuned ModularStarEncoder with a batch size of 2048 contrastive samples for 20,000 training steps.
66
+ The pre-training and fine-tuning were conducted on 512 NVIDIA Ampere (64GB) GPUs using the [Leonardo](https://arxiv.org/abs/2307.16885) supercomputer, requiring 450,000 GPU working hours.
67
+
68
+ | Hyperparameter | Value |
69
+ |--------------------------|-----------|
70
+ | Hidden size | 1024 |
71
+ | Max. position embeddings | 2048 |
72
+ | Num. of attention heads | 12 |
73
+ | Num. of key values heads | 4 |
74
+ | Num. of hidden layers | 36 |
75
+ | Attention | GQA |
76
+ | Num. of parameters | ≈1B |
77
+ |Loss function |CLIP loss |
78
+ |Multi-layer loss | yes |
79
+
80
+ ## Licence
81
+ The model is licensed under the BigCode OpenRAIL-M v1 license agreement. You can find the full agreement [here](https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement).