Update README.md

README.md

@@ -1,7 +1,5 @@
 # Model Card for climategpt/climategpt-70b
 - This model is the 70B parameter variant of the ClimateGPT model release.
-- Starting from Llama2 70B weights, the model undergoes continued pretraining and instruction finetuning on climate data.
-- The model is capable of answering questions and following instructions, especially tailored for the climate domain.
 
 ## Overview
 - **Developed by:** AppTek, Eqtylab, Erasmus AI
@@ -30,13 +28,7 @@ For example, it is possible to evaluate the model with https://github.com/Eleuth
 ## Training
 - For the Llama2 training data, we refer the user to https://huggingface.co/meta-llama/Llama-2-70b-chat-hf.
 - For continued pretraining, 4.2B climate domain tokens (tokenized by the Llama tokenizer) are used.
-- For instruction finetuning, about 579K instruction-completion pairs (both in the climate domain but also general domain) are used.
-
-## Evaluation
-Automatic evaluation is done via https://github.com/EleutherAI/lm-evaluation-harness, into which we also implemented custom evaluation tasks.
-TO BE ADDED
-We also perform human evaluation with experts in the climate domain.
-TO BE ADDED
+- For instruction finetuning, about 1.1B instruction-finetuning tokens (both in the climate domain but also general domain) are used.
 
 ## Environmental Impact
 - **Hardware Type:** H100