tricktreat
commited on
Commit
•
58bc3b2
1
Parent(s):
72f0347
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,29 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
datasets:
|
4 |
+
- timdettmers/openassistant-guanaco
|
5 |
+
library_name: adapter-transformers
|
6 |
+
---
|
7 |
+
|
8 |
+
## Hyperparameter
|
9 |
+
|
10 |
+
```bash
|
11 |
+
deepspeed --include localhost:0,1,2,3 sft.py --deepspeed dp_zero3.json \
|
12 |
+
--model_name_or_path="/home/shenyl/cached_models/meta-llama/Llama-2-7b-chat-hf" \
|
13 |
+
--dataset_name="timdettmers/openassistant-guanaco" \
|
14 |
+
--dataset_text_field="text" \
|
15 |
+
--report_to="tensorboard" \
|
16 |
+
--learning_rate=1e-5 \
|
17 |
+
--per_device_train_batch_size=6 \
|
18 |
+
--gradient_accumulation_steps=8 \
|
19 |
+
--output_dir="guanaco_Llama-2-7b-chat-hf" \
|
20 |
+
--logging_steps=1 \
|
21 |
+
--num_train_epochs=15 \
|
22 |
+
--max_steps=-1 \
|
23 |
+
--gradient_checkpointing \
|
24 |
+
--save_steps=0.3
|
25 |
+
```
|
26 |
+
|
27 |
+
## Dataset
|
28 |
+
|
29 |
+
`timdettmers/openassistant-guanaco`
|