Upload README.md with huggingface_hub

README.md

@@ -14,27 +14,27 @@ This folder contains `.cellm` model artifacts tested with the Cellm Rust CLI.
 ## Models
 
 ### Qwen2.5 0.5B Instruct (INT8)
-- **Path**: `../models/qwen2.5-0.5b-int8-v1.cellm`
+- **Path**: `models/qwen2.5-0.5b-int8-v1.cellm`
 - **Size**: ~472 MB
-- **Tokenizer**: `../models/qwen2.5-0.5b-bnb4/tokenizer.json`
+- **Tokenizer**: `models/qwen2.5-0.5b-bnb4/tokenizer.json`
 - **Type**: INT8 symmetric weight-only
 
 ### Gemma-3 1B IT (INT4, smallest)
-- **Path**: `../models/gemma-3-1b-it-int4-v1.cellm`
+- **Path**: `models/gemma-3-1b-it-int4-v1.cellm`
 - **Size**: ~478 MB
-- **Tokenizer**: `../models/hf/gemma-3-1b-it-full/tokenizer.json`
+- **Tokenizer**: `models/hf/gemma-3-1b-it-full/tokenizer.json`
 - **Type**: INT4 symmetric weight-only
 
 ### Gemma-3 1B IT (Mixed INT4, recommended)
-- **Path**: `../models/gemma-3-1b-it-mixed-int4-v1.cellm`
+- **Path**: `models/gemma-3-1b-it-mixed-int4-v1.cellm`
 - **Size**: ~1.0 GB
-- **Tokenizer**: `../models/hf/gemma-3-1b-it-full/tokenizer.json`
+- **Tokenizer**: `models/hf/gemma-3-1b-it-full/tokenizer.json`
 - **Type**: Mixed precision (attention/embeddings higher precision, MLP mostly INT4)
 
 ### Gemma-3 1B IT (INT8, most stable)
-- **Path**: `../models/gemma-3-1b-it-int8-v1.cellm`
+- **Path**: `models/gemma-3-1b-it-int8-v1.cellm`
 - **Size**: ~1.2 GB
-- **Tokenizer**: `../models/hf/gemma-3-1b-it-full/tokenizer.json`
+- **Tokenizer**: `models/hf/gemma-3-1b-it-full/tokenizer.json`
 - **Type**: INT8 symmetric weight-only
 
 ## Usage
@@ -71,3 +71,4 @@ Cellm is a Rust-native inference runtime focused on mobile/desktop local LLM ser
 
 ## License
 Please follow each upstream model license (Qwen and Gemma terms) when redistributing weights and tokenizers.
+