Update README.md
Browse files
README.md
CHANGED
|
@@ -2,11 +2,16 @@
|
|
| 2 |
license: llama3.1
|
| 3 |
base_model:
|
| 4 |
- THUDM/LongCite-llama3.1-8b
|
|
|
|
|
|
|
|
|
|
| 5 |
---
|
| 6 |
GGUF version of longcite, you need to add the following tokens as stop tokens : `[128000, 128007, 128009]` or `["<|begin_of_text|>", "<|end_header_id|>", "<|eot_id|>"]`
|
| 7 |
|
| 8 |
Be default, and it seems to be working so far, EOS token is 128007 (end_header_id). Working for citation and naive question-answer mode.
|
| 9 |
|
|
|
|
|
|
|
| 10 |
Example code
|
| 11 |
```python
|
| 12 |
|
|
|
|
| 2 |
license: llama3.1
|
| 3 |
base_model:
|
| 4 |
- THUDM/LongCite-llama3.1-8b
|
| 5 |
+
datasets:
|
| 6 |
+
- THUDM/LongCite-45k
|
| 7 |
+
pipeline_tag: text-generation
|
| 8 |
---
|
| 9 |
GGUF version of longcite, you need to add the following tokens as stop tokens : `[128000, 128007, 128009]` or `["<|begin_of_text|>", "<|end_header_id|>", "<|eot_id|>"]`
|
| 10 |
|
| 11 |
Be default, and it seems to be working so far, EOS token is 128007 (end_header_id). Working for citation and naive question-answer mode.
|
| 12 |
|
| 13 |
+
Not chat template provided as it requires python pre-processing (before being sent to LLM) and post-processing.
|
| 14 |
+
|
| 15 |
Example code
|
| 16 |
```python
|
| 17 |
|