File size: 3,492 Bytes
5d268dc |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 |
---
license: apache-2.0
base_model: microsoft/phi-2
tags:
- peft
- lora
- gsm8k
- math
- reasoning
- curriculum-learning
datasets:
- gsm8k
metrics:
- accuracy
library_name: peft
---
# phi-2 LoRA Adapter for GSM8K
This is a LoRA adapter for [microsoft/phi-2](https://huggingface.co/microsoft/phi-2) fine-tuned on the [GSM8K dataset](https://huggingface.co/datasets/gsm8k) for mathematical reasoning.
## Model Description
- **Base Model:** microsoft/phi-2
- **Training Method:** Curriculum learning - Complexity Score
- **Dataset:** GSM8K (Grade School Math 8K)
- **Task:** Mathematical word problem solving
- **Exact Match Accuracy:** 62.50%
## Training Details
LoRA adapter for PHI-2 trained with curriculum learning (complexity score method)
### Training Configuration
- **Method:** LoRA (Low-Rank Adaptation)
- **Rank:** 16
- **Alpha:** 32
- **Target Modules:** q_proj, k_proj, v_proj, o_proj
- **Dropout:** 0.1
- **Epochs:** 3
- **Batch Size:** 4 (with gradient accumulation of 4)
- **Learning Rate:** 3e-4
### Curriculum Learning
This model was trained using curriculum learning, where the model is exposed to progressively harder problems:
1. **Easy Stage:** Simple problems with fewer steps
2. **Normal Stage:** Moderate complexity problems
3. **Difficult Stage:** Complex multi-step problems
The curriculum was determined based on problem complexity (number of solution steps × operation complexity).
## Usage
### Loading the Adapter
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
# Load base model
base_model = AutoModelForCausalLM.from_pretrained(
"microsoft/phi-2",
device_map="auto",
torch_dtype="auto"
)
# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-2")
# Load LoRA adapter
model = PeftModel.from_pretrained(base_model, "CrystalRaindropsFall/phi2-gsm8k-curriculum-complexity")
# Inference
prompt = "Question: Janet's ducks lay 16 eggs per day. She eats three for breakfast every morning and bakes muffins for her friends every day with four. She sells the remainder at the farmers' market daily for $2 per fresh duck egg. How much in dollars does she make every day at the farmers' market?\nAnswer:"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```
### Using with Pipeline
```python
from transformers import pipeline
from peft import PeftModel, AutoPeftModelForCausalLM
# Load model with adapter
model = AutoPeftModelForCausalLM.from_pretrained(
"YOUR_USERNAME/REPO_NAME",
device_map="auto"
)
# Create pipeline
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
# Generate
result = pipe("Question: A robe takes 2 bolts of blue fiber and half that much white fiber. How many bolts in total does it take?\nAnswer:")
print(result[0]['generated_text'])
```
## Performance
Evaluated on GSM8K test set (512 samples):
| Metric | Score |
|--------|-------|
| Exact Match | 62.50% |
| Format Correct | 100% |
## Limitations
- Trained on grade school level math problems
- May struggle with problems requiring external knowledge
- Performance depends on problem complexity and wording
- Best used with base model's standard generation settings
## Acknowledgments
- Base model: microsoft/phi-2
- Dataset: GSM8K by Cobbe et al.
- Training framework: HuggingFace PEFT
## License
Apache 2.0 (following base model license)
|