Update README.md
Browse files
README.md
CHANGED
|
@@ -77,7 +77,7 @@ print(tokenizer.decode(outputs[0][inputs.input_ids.shape[-1]:], skip_special_tok
|
|
| 77 |
## ⚡ vLLM
|
| 78 |
|
| 79 |
```
|
| 80 |
-
vllm serve
|
| 81 |
```
|
| 82 |
|
| 83 |
```python
|
|
@@ -108,22 +108,6 @@ print(ask("Find the degree for the given field extension Q(sqrt(2), sqrt(3), sqr
|
|
| 108 |
|
| 109 |
---
|
| 110 |
|
| 111 |
-
## 🔧 Training Details
|
| 112 |
-
|
| 113 |
-
Base Model Qwen/Qwen3.5‑2B
|
| 114 |
-
Method LoRA (Low‑Rank Adaptation)
|
| 115 |
-
LoRA Rank / Alpha r=16, alpha=32
|
| 116 |
-
Dataset Size 10,000 samples (2,000 from each of 5 datasets)
|
| 117 |
-
Training Epochs 1
|
| 118 |
-
Max Sequence Length 512 tokens
|
| 119 |
-
Batch Size 1 (with gradient accumulation ×8)
|
| 120 |
-
Learning Rate 5e‑5
|
| 121 |
-
Precision FP16
|
| 122 |
-
Framework Transformers + PEFT
|
| 123 |
-
Hardware Single NVIDIA T4 16GB (Kaggle)
|
| 124 |
-
|
| 125 |
-
---
|
| 126 |
-
|
| 127 |
## 🏗️ Architecture
|
| 128 |
|
| 129 |
OmniMath‑2B fully preserves Qwen3.5‑2B's design:
|
|
|
|
| 77 |
## ⚡ vLLM
|
| 78 |
|
| 79 |
```
|
| 80 |
+
vllm serve ZirTech/OmniMath-2B --tensor-parallel-size 1 --max-model-len 4096
|
| 81 |
```
|
| 82 |
|
| 83 |
```python
|
|
|
|
| 108 |
|
| 109 |
---
|
| 110 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 111 |
## 🏗️ Architecture
|
| 112 |
|
| 113 |
OmniMath‑2B fully preserves Qwen3.5‑2B's design:
|