Downloads · 30 days
5
56% of all-time downloads
TML_Trained_model_random_dataset
aaryankamdar/TML_Trained_model_random_dataset
TML_Trained_model_random_dataset is a machine learning model from aaryankamdar. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Base Model: Qwen2.5-3B-Instruct Fine-tuning Method: LoRA (Low-Rank Adaptation) Dataset: Random15k.json (15,000 randomly sampled examples from AceReason-100K) Hardware: NVIDIA A100 (80 GB, Normal Queue)
All-time downloads
9
Public
Repo size
5 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
Other5 GB · 100%
At a glance
- Model type
- qwen2
- Access
- Public
- Created
- Oct 10, 2025
- Updated
- Oct 11, 2025
- SHA
- 5161f10f
Model card
From the Hugging Face model README
Base Model: Qwen2.5-3B-Instruct Fine-tuning Method: LoRA (Low-Rank Adaptation) Dataset: Random_15k.json (15,000 randomly sampled examples from AceReason-100K) Hardware: NVIDIA A100 (80 GB, Normal Queue)
Objective: Establish a baseline reasoning model trained on a randomly sampled dataset to measure the effect of data curation on performance.
Training Setup: • Epochs: 3 • Learning Rate: 2e-4 • Batch Size: 2 (grad accum: 8, effective 16) • Optimizer: AdamW • Scheduler: Cosine • Warmup Ratio: 0.03 • Precision: BFloat16 • Sequence Length: 4096 • Gradient Checkpointing: Enabled
LoRA Parameters: • r = 8 • α = 32 • Dropout = 0.05 • Target: all
Results: • MATH-500 Accuracy: ~30.0% • Serves as baseline for comparison with curated training
- Type
- qwen2
- Created
- Oct 10, 2025
- Updated
- Oct 11, 2025