Update README.md
Browse files
README.md
CHANGED
|
@@ -15,7 +15,8 @@ contains less than 100 tokens). The maximum token size for Orca2 is 4096 so a si
|
|
| 15 |
(considering propt instructions) has been used. Chunking did not consider context (text data might split within the context).
|
| 16 |
Evaluation set has been generated similar method on 1% of the raw data with LLama2 chat (https://huggingface.co/TheBloke/Llama-2-13B-chat-GGUF).
|
| 17 |
|
| 18 |
-
Trained locally on 2x3090 GPU with vanila DDP with HuggingFace Accelerate for 50 Epoch.
|
|
|
|
| 19 |
|
| 20 |
|
| 21 |
Chat with model sample code:
|
|
|
|
| 15 |
(considering propt instructions) has been used. Chunking did not consider context (text data might split within the context).
|
| 16 |
Evaluation set has been generated similar method on 1% of the raw data with LLama2 chat (https://huggingface.co/TheBloke/Llama-2-13B-chat-GGUF).
|
| 17 |
|
| 18 |
+
Trained locally on 2x3090 GPU with vanila DDP with HuggingFace Accelerate for 50 Epoch.
|
| 19 |
+
As I wanted to add new knowledge to the base model r=128 and lora_alpha=128 has been used -> LoRA weights are 3.5% of the base model.
|
| 20 |
|
| 21 |
|
| 22 |
Chat with model sample code:
|