Shell2334 commited on
Commit
be970ff
·
verified ·
1 Parent(s): 5a2885e

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +112 -131
README.md CHANGED
@@ -1,206 +1,187 @@
 
1
  ---
 
2
  base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
3
  library_name: peft
 
 
 
4
  tags:
5
- - base_model:adapter:TinyLlama/TinyLlama-1.1B-Chat-v1.0
6
- - lora
7
- - transformers
 
8
  ---
9
 
10
- # Model Card for Model ID
11
-
12
- <!-- Provide a quick summary of what the model is/does. -->
13
-
14
-
15
-
16
- ## Model Details
17
-
18
- ### Model Description
19
-
20
- <!-- Provide a longer summary of what this model is. -->
21
-
22
-
23
-
24
- - **Developed by:** [More Information Needed]
25
- - **Funded by [optional]:** [More Information Needed]
26
- - **Shared by [optional]:** [More Information Needed]
27
- - **Model type:** [More Information Needed]
28
- - **Language(s) (NLP):** [More Information Needed]
29
- - **License:** [More Information Needed]
30
- - **Finetuned from model [optional]:** [More Information Needed]
31
-
32
- ### Model Sources [optional]
33
-
34
- <!-- Provide the basic links for the model. -->
35
-
36
- - **Repository:** [More Information Needed]
37
- - **Paper [optional]:** [More Information Needed]
38
- - **Demo [optional]:** [More Information Needed]
39
-
40
- ## Uses
41
-
42
- <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
43
-
44
- ### Direct Use
45
-
46
- <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
47
-
48
- [More Information Needed]
49
-
50
- ### Downstream Use [optional]
51
-
52
- <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
53
-
54
- [More Information Needed]
55
-
56
- ### Out-of-Scope Use
57
-
58
- <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
59
-
60
- [More Information Needed]
61
-
62
- ## Bias, Risks, and Limitations
63
-
64
- <!-- This section is meant to convey both technical and sociotechnical limitations. -->
65
-
66
- [More Information Needed]
67
 
68
- ### Recommendations
69
 
70
- <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
71
 
72
- Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
73
 
74
- ## How to Get Started with the Model
75
 
76
- Use the code below to get started with the model.
 
 
 
 
 
 
77
 
78
- [More Information Needed]
79
 
80
- ## Training Details
81
 
82
- ### Training Data
83
 
84
- <!-- 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. -->
 
 
 
85
 
86
- [More Information Needed]
87
 
88
- ### Training Procedure
89
 
90
- <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
91
 
92
- #### Preprocessing [optional]
 
 
 
 
 
 
 
 
 
93
 
94
- [More Information Needed]
95
 
 
96
 
97
- #### Training Hyperparameters
98
 
99
- - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
100
 
101
- #### Speeds, Sizes, Times [optional]
102
 
103
- <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
 
 
 
 
104
 
105
- [More Information Needed]
106
 
107
- ## Evaluation
108
 
109
- <!-- This section describes the evaluation protocols and provides the results. -->
110
 
111
- ### Testing Data, Factors & Metrics
112
 
113
- #### Testing Data
114
 
115
- <!-- This should link to a Dataset Card if possible. -->
116
 
117
- [More Information Needed]
118
 
119
- #### Factors
120
 
121
- <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
122
 
123
- [More Information Needed]
124
 
125
- #### Metrics
 
 
 
 
126
 
127
- <!-- These are the evaluation metrics being used, ideally with a description of why. -->
128
 
129
- [More Information Needed]
130
 
131
- ### Results
 
 
 
 
132
 
133
- [More Information Needed]
134
 
135
- #### Summary
136
 
 
137
 
 
138
 
139
- ## Model Examination [optional]
140
 
141
- <!-- Relevant interpretability work for the model goes here -->
142
 
143
- [More Information Needed]
 
 
 
144
 
145
- ## Environmental Impact
146
 
147
- <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
148
 
149
- 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).
150
 
151
- - **Hardware Type:** [More Information Needed]
152
- - **Hours used:** [More Information Needed]
153
- - **Cloud Provider:** [More Information Needed]
154
- - **Compute Region:** [More Information Needed]
155
- - **Carbon Emitted:** [More Information Needed]
156
 
157
- ## Technical Specifications [optional]
158
 
159
- ### Model Architecture and Objective
160
 
161
- [More Information Needed]
162
 
163
- ### Compute Infrastructure
164
 
165
- [More Information Needed]
166
 
167
- #### Hardware
168
 
169
- [More Information Needed]
170
 
171
- #### Software
172
 
173
- [More Information Needed]
174
 
175
- ## Citation [optional]
176
 
177
- <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
178
 
179
- **BibTeX:**
180
 
181
- [More Information Needed]
182
 
183
- **APA:**
184
 
185
- [More Information Needed]
186
 
187
- ## Glossary [optional]
188
 
189
- <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
190
 
191
- [More Information Needed]
192
 
193
- ## More Information [optional]
194
 
195
- [More Information Needed]
196
 
197
- ## Model Card Authors [optional]
198
 
199
- [More Information Needed]
200
 
201
- ## Model Card Contact
202
 
203
- [More Information Needed]
204
- ### Framework versions
205
 
206
- - PEFT 0.21.2
 
1
+
2
  ---
3
+ license: apache-2.0
4
  base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
5
  library_name: peft
6
+ pipeline_tag: text-classification
7
+ language:
8
+ - en
9
  tags:
10
+ - peft
11
+ - lora
12
+ - text-classification
13
+ - travel
14
  ---
15
 
16
+ # INFO 642 - Travel Request Classifier (TinyLlama + LoRA)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
17
 
18
+ ## What the Model Does
19
 
20
+ This model is a seven-category text classifier developed for our INFO 642 Modular AI Travel Planner project.
21
 
22
+ The model identifies the user's travel request and helps the n8n workflow select the appropriate module.
23
 
24
+ It predicts one of seven labels:
25
 
26
+ 1. NEW_TRIP - Create a new travel itinerary.
27
+ 2. CHANGE_FLIGHT - Modify an existing flight.
28
+ 3. CHANGE_HOTEL - Modify an existing hotel.
29
+ 4. CHANGE_ATTRACTION - Modify an attraction or activity.
30
+ 5. CHANGE_RESTAURANT - Modify a restaurant or meal.
31
+ 6. MULTI_CHANGE - Modify multiple itinerary components.
32
+ 7. OTHER_OR_UNCLEAR - Handle unclear, general, or unsupported requests.
33
 
34
+ In our workflow, this model is intended to replace the original Groq-based placeholder classifier. The other modules handle itinerary generation, data lookup, budget calculations, and itinerary modifications.
35
 
36
+ The classifier does not make reservations or payments.
37
 
38
+ ## Base Model
39
 
40
+ - Base model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
41
+ - Model type: Seven-class text classification
42
+ - Fine-tuning method: Low-Rank Adaptation (LoRA)
43
+ - Framework: Hugging Face Transformers and PEFT
44
 
45
+ We followed the Lesson 16 TinyLlama training approach and added a new seven-class classification head.
46
 
47
+ LoRA allows us to train a small number of parameters instead of updating the entire base model.
48
 
49
+ Training configuration:
50
 
51
+ | Parameter | Value |
52
+ |---|---|
53
+ | LoRA rank | 8 |
54
+ | LoRA alpha | 16 |
55
+ | LoRA dropout | 0.05 |
56
+ | Target modules | q_proj, k_proj, v_proj, o_proj |
57
+ | Epochs | 3 |
58
+ | Batch size | 8 |
59
+ | Learning rate | 0.0002 |
60
+ | Trainable parameters | 2,267,136 |
61
 
62
+ Only approximately 0.22% of the model's parameters were trainable.
63
 
64
+ ## Training Data
65
 
66
+ We created a synthetic English dataset of 280 travel-related user requests for this course project.
67
 
68
+ The dataset includes seven balanced categories, with 40 examples per category.
69
 
70
+ Data split:
71
 
72
+ | Dataset | Samples |
73
+ |---|---:|
74
+ | Training set | 224 |
75
+ | Held-out test set | 56 |
76
+ | Total | 280 |
77
 
78
+ Each category contains 32 training examples and 8 held-out test examples.
79
 
80
+ The training data covers requests such as creating trips, changing flights, finding cheaper hotels, replacing attractions, modifying restaurants, and changing multiple itinerary components.
81
 
82
+ Some examples also include unclear requests and unsupported booking or payment requests.
83
 
84
+ The test data was not used for model training.
85
 
86
+ All examples are synthetic and do not represent real customer conversations.
87
 
88
+ ## Results
89
 
90
+ We evaluated the fine-tuned model using 56 held-out test requests.
91
 
92
+ ### Overall Accuracy
93
 
94
+ **Held-out test accuracy: 98.2% (55/56 correct)**
95
 
96
+ Additional model comparisons:
97
 
98
+ | Model | Accuracy |
99
+ |---|---:|
100
+ | BART Zero-shot | 46.4% |
101
+ | TinyLlama with untrained classification head | 14.3% |
102
+ | TinyLlama + LoRA | 98.2% |
103
 
104
+ The fine-tuned model performed substantially better on our synthetic test set.
105
 
106
+ ### Training Loss
107
 
108
+ | Epoch | Loss |
109
+ |---|---:|
110
+ | 1 | 1.944 |
111
+ | 2 | 0.461 |
112
+ | 3 | 0.091 |
113
 
114
+ The training loss decreased across all three epochs.
115
 
116
+ ### Confusion Matrix Analysis
117
 
118
+ The fine-tuned model correctly classified 55 out of 56 test examples.
119
 
120
+ Six categories achieved perfect recall.
121
 
122
+ The only classification error occurred when a MULTI_CHANGE request was predicted as CHANGE_ATTRACTION.
123
 
124
+ The relevant category metrics were:
125
 
126
+ | Category | Precision | Recall | F1-score |
127
+ |---|---:|---:|---:|
128
+ | CHANGE_ATTRACTION | 0.89 | 1.00 | 0.94 |
129
+ | MULTI_CHANGE | 1.00 | 0.88 | 0.93 |
130
 
131
+ The macro-average F1-score was 0.98.
132
 
133
+ This indicates that requests involving multiple changes may sometimes be confused with requests involving only one attraction change.
134
 
135
+ The complete confusion matrix and classification report are included in the training notebook.
136
 
137
+ ## Limitations
 
 
 
 
138
 
139
+ Although the model achieved high accuracy on the held-out dataset, several limitations remain.
140
 
141
+ 1. **Synthetic training data:** The dataset contains only 280 synthetic examples. Real user requests may be more complex or expressed differently.
142
 
143
+ 2. **Multi-change confusion:** Requests involving multiple itinerary components may sometimes be classified as a single-category change.
144
 
145
+ 3. **Ambiguous language:** Short or unclear requests may not provide enough information for reliable classification.
146
 
147
+ 4. **Limited task coverage:** The model only supports the seven predefined categories. It cannot independently create itineraries, verify live availability, make reservations, or process payments.
148
 
149
+ 5. **Generalization:** The 98.2% accuracy was measured on a small synthetic held-out dataset. Performance on real travel conversations has not been established.
150
 
151
+ The model is intended for educational workflow routing and should not be used as an independent booking or payment system.
152
 
153
+ ## Fresh-Session Model Verification
154
 
155
+ After training, we uploaded the LoRA adapter, classification head, tokenizer, and label configuration to this public Hugging Face repository.
156
 
157
+ We then opened a fresh Google Colab notebook and loaded the model directly from Hugging Face.
158
 
159
+ Test input:
160
 
161
+ > Please update the afternoon sightseeing activity to the science center.
162
 
163
+ Expected prediction from the original training notebook:
164
 
165
+ `CHANGE_ATTRACTION`
166
 
167
+ Prediction after loading from Hugging Face:
168
 
169
+ `CHANGE_ATTRACTION`
170
 
171
+ **Result: PASS - The fresh-session prediction matched the original notebook.**
172
 
173
+ This verifies that the uploaded model can be loaded in a separate session and reproduce the prediction for this test example.
174
 
175
+ ### Fresh-Session Test Screenshot
176
 
177
+ ![Fresh-session model loading and prediction result](fresh_session_prediction.png)
178
 
179
+ ## Project Information
180
 
181
+ Course: INFO 642
182
 
183
+ Project: Modular AI Travel Planner
184
 
185
+ Model repository: Shell2334/info642-travel-classifier-lora
 
186
 
187
+ Purpose: Educational research and demonstration.