midwestern-simulation-active
/

essence-3b-v1.1

Model card Files Files and versions

metalure commited on Sep 12, 2025

Commit

6b4d420

·

verified ·

1 Parent(s): f630132

Update README.md

Files changed (1) hide show

README.md +2 -2

README.md CHANGED Viewed

@@ -6,7 +6,7 @@ base_model:
 ---
-# Essence 3B V1.1 alpha
 This is a system using two versions of SmolLM3-3B-Base, the 'encoder', is finetuned to turn a text into a set of embedding tokens which can be reconstituted back into the original text by the decoder. In addition to “vanilla” reconstruction, this model was trained for span-corruption and masked language modelling.
@@ -26,7 +26,7 @@ from huggingface_hub import hf_hub_download
 device = torch.device("cuda:0")
 dtype = torch.bfloat16
 base_model_id = "HuggingFaceTB/SmolLM3-3B-Base"
-compressor_id = "midwestern-simulation/essence-3b-v1.1-alpha"
 # === MODEL LOADING ===

 ---
+# Essence 3B V1.1
 This is a system using two versions of SmolLM3-3B-Base, the 'encoder', is finetuned to turn a text into a set of embedding tokens which can be reconstituted back into the original text by the decoder. In addition to “vanilla” reconstruction, this model was trained for span-corruption and masked language modelling.
 device = torch.device("cuda:0")
 dtype = torch.bfloat16
 base_model_id = "HuggingFaceTB/SmolLM3-3B-Base"
+compressor_id = "midwestern-simulation/essence-3b-v1.1"
 # === MODEL LOADING ===