Update README.md
Browse files
README.md
CHANGED
|
@@ -9,7 +9,7 @@ Replete-LLM-Qwen2-7b_Beta-Preview
|
|
| 9 |
|
| 10 |
This is a preview look at our flagship model, Replete-LLM. This version of the model has only been trained for 1 epoch on the dataset (Linked bellow). The final model will be trained on a full 5 epochs using Qlora and Unsloth.
|
| 11 |
|
| 12 |
-
- Replete-AI/Everything_Instruct_8k_context_filtered
|
| 13 |
|
| 14 |
# Model card:
|
| 15 |
|
|
@@ -78,6 +78,7 @@ The code to finetune this AI model can be found bellow
|
|
| 78 |
!export PATH=/usr/local/cuda/bin:$PATH
|
| 79 |
!export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
|
| 80 |
!export NCCL_P2P_LEVEL=NVL
|
|
|
|
| 81 |
!export NCCL_DEBUG=INFO
|
| 82 |
!export NCCL_DEBUG_SUBSYS=ALL
|
| 83 |
!export TORCH_DISTRIBUTED_DEBUG=INFO
|
|
|
|
| 9 |
|
| 10 |
This is a preview look at our flagship model, Replete-LLM. This version of the model has only been trained for 1 epoch on the dataset (Linked bellow). The final model will be trained on a full 5 epochs using Qlora and Unsloth.
|
| 11 |
|
| 12 |
+
- https://huggingface.co/datasets/Replete-AI/Everything_Instruct_8k_context_filtered
|
| 13 |
|
| 14 |
# Model card:
|
| 15 |
|
|
|
|
| 78 |
!export PATH=/usr/local/cuda/bin:$PATH
|
| 79 |
!export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
|
| 80 |
!export NCCL_P2P_LEVEL=NVL
|
| 81 |
+
|
| 82 |
!export NCCL_DEBUG=INFO
|
| 83 |
!export NCCL_DEBUG_SUBSYS=ALL
|
| 84 |
!export TORCH_DISTRIBUTED_DEBUG=INFO
|