Update README.md

README.md

@@ -4,13 +4,13 @@ library_name: femr
 tags:
 - healthcare
 ---
-# CLMBR-T-Base-Synthetic
+# CLMBR-T-Base-Random
 
-This is a CLMBR model that was "pretrained" on **synthetic data**. The purpose of this model is to test code pipelines and demonstrate how to use CLMBR before applying for access to the official CLMBR release that was trained on real Stanford Hospital data.
+This is a CLMBR model that was randomly initialized with a dummy vocabulary. The purpose of this model is to test code pipelines and demonstrate how to use CLMBR before applying for access to the official CLMBR release that was trained on real Stanford Hospital data.
 
 The model architecture is CLMBR-T-Base (144M params), as originally described in [the EHRSHOT paper (Wornow et al. 2023)](https://arxiv.org/abs/2307.02028), and based on the architecture originally developed in [the Clinical Language Modeling Based Representations paper (Steinberg et al. 2021)](https://www.sciencedirect.com/science/article/pii/S1532046420302653)
 
-The synthetic data used for training is uniformly sampled random data with no intrinsic signal, so **this model has no clinical or research use.**
+The weights are random, so **this model has no clinical or research use.**
 
 ## Model Details
 
@@ -35,7 +35,7 @@ This model generates dense representations for patients based on the structured
 
 These representations can then be used for downstream tasks such as predicting diagnoses, detecting anomalies, or doing propensity score matching for causal inference.
 
-However, please note that **this version of the model was trained on synthetic data.** Thus, the outputs will be random/meaningless.
+However, please note that **this version of the model has random weights.** Thus, the outputs will be random/meaningless.
 
 
 ## How to Get Started with the Model
@@ -93,11 +93,11 @@ with torch.no_grad():
 
 ## Training Details
 
-The synthetic data used for training is uniformly sampled random data with no intrinsic signal, so this model itself has no clinical or research use. 
+This model is not trained.
 
 ## Evaluation
 
-As this model was trained on random synthetic data, we expect it to perform no better than chance.
+None, as the weights are random.
 
 ## Technical Specifications
 
@@ -105,7 +105,7 @@ Please see [the EHRSHOT paper (Wornow et al. 2023)](https://arxiv.org/abs/2307.0
 
 ### Compute Infrastructure
 
-This model was trained on a single V100 GPU.
+This model was not trained.
 
 #### Software