Update README.md
Browse files
README.md
CHANGED
|
@@ -76,6 +76,18 @@ outputs = model.generate(
|
|
| 76 |
input_len=len(inputs["input_ids"])
|
| 77 |
print(tokenizer.decode(outputs[0][input_len:]))
|
| 78 |
```
|
| 79 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
# More information
|
| 81 |
For further information, refer to [StarCoder](https://huggingface.co/bigcode/starcoder).
|
|
|
|
| 76 |
input_len=len(inputs["input_ids"])
|
| 77 |
print(tokenizer.decode(outputs[0][input_len:]))
|
| 78 |
```
|
| 79 |
+
# Updates
|
| 80 |
+
Gradio dataset `.filter(lambda x : ("gradio" in x["content"] or "gr." in x["content"]) and "streamlit" not in x["content"]`)
|
| 81 |
+
Guanaco `ArmelR/oasst1_guanaco`
|
| 82 |
+
- StarCoderbase (950, 1350)
|
| 83 |
+
- max_steps = 2000
|
| 84 |
+
- shuffle_buffer = 100
|
| 85 |
+
- batch_size = 2
|
| 86 |
+
- gradient_accumulation_steps = 4
|
| 87 |
+
- num_warmup_steps = 100
|
| 88 |
+
- weight_decay = 0.01
|
| 89 |
+
- StarCoderplus (2000)
|
| 90 |
+
|
| 91 |
+
Guanaco multi-turn (HuggingFaceH4/oasst1_en)
|
| 92 |
# More information
|
| 93 |
For further information, refer to [StarCoder](https://huggingface.co/bigcode/starcoder).
|