haielab commited on
Commit
4258da5
·
verified ·
1 Parent(s): a00009d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +26 -0
README.md CHANGED
@@ -39,8 +39,34 @@ LoRA rank-16 adapter fine-tuned from **`kfdong/STP_model_Lean_0320`** to assist
39
  | **Context length** | 1792 tokens |
40
  | **Hardware** | 1 × H100 80 GB |
41
 
 
42
  ---
43
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
44
  ### Results (1 epoch, STP Lean corpus)
45
 
46
  | Metric | Value |
 
39
  | **Context length** | 1792 tokens |
40
  | **Hardware** | 1 × H100 80 GB |
41
 
42
+
43
  ---
44
 
45
+ #### Training Hyperparameters
46
+
47
+ | Setting | Value |
48
+ |-------------------------------|---------------------------------------------|
49
+ | Precision / regime | **bf16 mixed precision** |
50
+ | Epochs | **1** |
51
+ | Max sequence length | 1 792 tokens (right-padding) |
52
+ | Per-device train batch size | 6 |
53
+ | Per-device eval batch size | 2 |
54
+ | Gradient accumulation steps | 1 (effective batch = 6) |
55
+ | Optimizer | AdamW |
56
+ | Learning rate schedule | **2 × 10⁻⁴** cosine, warm-up **3 %** |
57
+ | Weight decay | 0.01 |
58
+ | LoRA rank / α / dropout | r = 16, α = 32 (2 × r), dropout = 0.05 |
59
+ | Gradient checkpointing | Enabled (memory-efficient) |
60
+ | Flash-Attention v2 | Enabled |
61
+ | Logging | every 50 steps |
62
+ | Evaluation strategy | once per epoch |
63
+ | Save strategy | once per epoch |
64
+ | Seed | 42 |
65
+ | Hardware | 1 × H100 80 GB |
66
+
67
+ ---
68
+
69
+
70
  ### Results (1 epoch, STP Lean corpus)
71
 
72
  | Metric | Value |