Mihaiii commited on
Commit
fab54d8
1 Parent(s): 864c283

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +38 -194
README.md CHANGED
@@ -12,197 +12,41 @@ language:
12
  license: apache-2.0
13
  ---
14
 
15
- # Model Card for Model ID
16
-
17
- <!-- Provide a quick summary of what the model is/does. -->
18
-
19
-
20
-
21
- ## Model Details
22
-
23
- ### Model Description
24
-
25
- <!-- Provide a longer summary of what this model is. -->
26
-
27
- This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
28
-
29
- - **Developed by:** [More Information Needed]
30
- - **Funded by [optional]:** [More Information Needed]
31
- - **Shared by [optional]:** [More Information Needed]
32
- - **Model type:** [More Information Needed]
33
- - **Language(s) (NLP):** [More Information Needed]
34
- - **License:** [More Information Needed]
35
- - **Finetuned from model [optional]:** [More Information Needed]
36
-
37
- ### Model Sources [optional]
38
-
39
- <!-- Provide the basic links for the model. -->
40
-
41
- - **Repository:** [More Information Needed]
42
- - **Paper [optional]:** [More Information Needed]
43
- - **Demo [optional]:** [More Information Needed]
44
-
45
- ## Uses
46
-
47
- <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
48
-
49
- ### Direct Use
50
-
51
- <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
52
-
53
- [More Information Needed]
54
-
55
- ### Downstream Use [optional]
56
-
57
- <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
58
-
59
- [More Information Needed]
60
-
61
- ### Out-of-Scope Use
62
-
63
- <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
64
-
65
- [More Information Needed]
66
-
67
- ## Bias, Risks, and Limitations
68
-
69
- <!-- This section is meant to convey both technical and sociotechnical limitations. -->
70
-
71
- [More Information Needed]
72
-
73
- ### Recommendations
74
-
75
- <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
76
-
77
- Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
78
-
79
- ## How to Get Started with the Model
80
-
81
- Use the code below to get started with the model.
82
-
83
- [More Information Needed]
84
-
85
- ## Training Details
86
-
87
- ### Training Data
88
-
89
- <!-- 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. -->
90
-
91
- [More Information Needed]
92
-
93
- ### Training Procedure
94
-
95
- <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
96
-
97
- #### Preprocessing [optional]
98
-
99
- [More Information Needed]
100
-
101
-
102
- #### Training Hyperparameters
103
-
104
- - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
105
-
106
- #### Speeds, Sizes, Times [optional]
107
-
108
- <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
109
-
110
- [More Information Needed]
111
-
112
- ## Evaluation
113
-
114
- <!-- This section describes the evaluation protocols and provides the results. -->
115
-
116
- ### Testing Data, Factors & Metrics
117
-
118
- #### Testing Data
119
-
120
- <!-- This should link to a Dataset Card if possible. -->
121
-
122
- [More Information Needed]
123
-
124
- #### Factors
125
-
126
- <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
127
-
128
- [More Information Needed]
129
-
130
- #### Metrics
131
-
132
- <!-- These are the evaluation metrics being used, ideally with a description of why. -->
133
-
134
- [More Information Needed]
135
-
136
- ### Results
137
-
138
- [More Information Needed]
139
-
140
- #### Summary
141
-
142
-
143
-
144
- ## Model Examination [optional]
145
-
146
- <!-- Relevant interpretability work for the model goes here -->
147
-
148
- [More Information Needed]
149
-
150
- ## Environmental Impact
151
-
152
- <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
153
-
154
- 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).
155
-
156
- - **Hardware Type:** [More Information Needed]
157
- - **Hours used:** [More Information Needed]
158
- - **Cloud Provider:** [More Information Needed]
159
- - **Compute Region:** [More Information Needed]
160
- - **Carbon Emitted:** [More Information Needed]
161
-
162
- ## Technical Specifications [optional]
163
-
164
- ### Model Architecture and Objective
165
-
166
- [More Information Needed]
167
-
168
- ### Compute Infrastructure
169
-
170
- [More Information Needed]
171
-
172
- #### Hardware
173
-
174
- [More Information Needed]
175
-
176
- #### Software
177
-
178
- [More Information Needed]
179
-
180
- ## Citation [optional]
181
-
182
- <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
183
-
184
- **BibTeX:**
185
-
186
- [More Information Needed]
187
-
188
- **APA:**
189
-
190
- [More Information Needed]
191
-
192
- ## Glossary [optional]
193
-
194
- <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
195
-
196
- [More Information Needed]
197
-
198
- ## More Information [optional]
199
-
200
- [More Information Needed]
201
-
202
- ## Model Card Authors [optional]
203
-
204
- [More Information Needed]
205
-
206
- ## Model Card Contact
207
-
208
- [More Information Needed]
 
12
  license: apache-2.0
13
  ---
14
 
15
+ This is pruned down version of [cognitivecomputations/dolphin-2.6-mistral-7b-dpo](https://huggingface.co/cognitivecomputations/dolphin-2.6-mistral-7b-dpo) from 7.24B params to 5.93B params (~ 82%).
16
+
17
+ # Steps to replicate:
18
+
19
+ Use [laserQlora.ipynb](https://github.com/cognitivecomputations/laserRMT/blob/main/laserQlora.ipynb) from [cognitivecomputations/laserRMT](https://github.com/cognitivecomputations/laserRMT) to determine which layers should be eliminated.
20
+
21
+ Replace `model_name = "mistralai/Mistral-7B-v0.1"` with `model_name = "cognitivecomputations/dolphin-2.6-mistral-7b-dpo"`.
22
+ I also ran the script only for `self_attn.v_proj` (so change the script to `layer_types=["self_attn.v_proj"]`)
23
+
24
+ Order by snr descending and eliminate top layers using [mergekit](https://github.com/arcee-ai/mergekit).
25
+ The threshold for elimination is up to you, depeding on how many layers you want removed. I decided to remove 6 layers (indexes: 3, 5, 16, 18, 19, 24 )
26
+
27
+ Here is the mergekit config:
28
+ ```yml
29
+ slices:
30
+ - sources:
31
+ - model: "cognitivecomputations/dolphin-2.6-mistral-7b-dpo"
32
+ layer_range: [0, 3]
33
+ - sources:
34
+ - model: "cognitivecomputations/dolphin-2.6-mistral-7b-dpo"
35
+ layer_range: [4, 5]
36
+ - sources:
37
+ - model: "cognitivecomputations/dolphin-2.6-mistral-7b-dpo"
38
+ layer_range: [6, 16]
39
+ - sources:
40
+ - model: "cognitivecomputations/dolphin-2.6-mistral-7b-dpo"
41
+ layer_range: [17, 18]
42
+ - sources:
43
+ - model: "cognitivecomputations/dolphin-2.6-mistral-7b-dpo"
44
+ layer_range: [20, 24]
45
+ - sources:
46
+ - model: "cognitivecomputations/dolphin-2.6-mistral-7b-dpo"
47
+ layer_range: [25, 32]
48
+ merge_method: passthrough
49
+ dtype: bfloat16
50
+ ```
51
+
52
+ The model outputted by mergekit with this configuration is this model (dolphin-2.6-mistral-7b-dpo-5.93B).