muditbaid commited on
Commit
a7b3160
·
verified ·
1 Parent(s): 82c45f1

README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -2
README.md CHANGED
@@ -8,8 +8,13 @@ tags:
8
  - qlora
9
  - peft
10
  - llama3
 
11
  base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
12
  pipeline_tag: text-classification
 
 
 
 
13
  ---
14
  # llama3.1-8b-Instruct-qlora-hatexplain
15
 
@@ -50,7 +55,7 @@ python scripts/qlora_inference.py \
50
  ## Training Notes
51
 
52
  - LLaMA-Factory SFT stage with LoRA rank 8, alpha 16, dropout 0.05.
53
- - Cutoff length 2048, cosine scheduler, 3 epochs, learning rate 5e-5.
54
  - QLoRA (4-bit) backbone for efficient fine-tuning on a single GPU.
55
 
56
- Refer to `config/llama31_hatexplain_qlora_sft.yaml` for the full set of hyperparameters.
 
8
  - qlora
9
  - peft
10
  - llama3
11
+ - llama-factory
12
  base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
13
  pipeline_tag: text-classification
14
+ datasets:
15
+ - Hate-speech-CNERG/hatexplain
16
+ metrics:
17
+ - accuracy
18
  ---
19
  # llama3.1-8b-Instruct-qlora-hatexplain
20
 
 
55
  ## Training Notes
56
 
57
  - LLaMA-Factory SFT stage with LoRA rank 8, alpha 16, dropout 0.05.
58
+ - Cutoff length 1024, cosine scheduler, 3 epochs, learning rate 2e-5.
59
  - QLoRA (4-bit) backbone for efficient fine-tuning on a single GPU.
60
 
61
+ Refer to `config/llama31_hatexplain_qlora_sft.yaml` for the full set of hyperparameters.