Update username to leochen085; fix base checkpoint download snippet

README.md

@@ -12,8 +12,8 @@ tags:
 base_model:
 - google/gemma-3-270m
 datasets:
-- LeoChen085/SlipDataset
-- LeoChen085/SlipSFTDataset
+- leochen085/SlipDataset
+- leochen085/SlipSFTDataset
 pipeline_tag: feature-extraction
 ---
 
@@ -25,9 +25,9 @@ Yuliang Chen, Arvind Pillai, Yu Yvonne Wu, Tess Z. Griffin, Lisa Marsch, Michael
 
 *Dartmouth College*
 
-[[Paper]](asset/manuscript.pdf) [[Code]](https://github.com/yuc0805/SLIP) [[Llama variant]](https://huggingface.co/LeoChen085/SLIP-Llama) [[Dataset]](https://huggingface.co/datasets/LeoChen085/SlipDataset) [[SFT Dataset]](https://huggingface.co/datasets/LeoChen085/SlipSFTDataset)
+[[Paper]](asset/manuscript.pdf) [[Code]](https://github.com/yuc0805/SLIP) [[Llama variant]](https://huggingface.co/leochen085/SLIP-Llama) [[Dataset]](https://huggingface.co/datasets/leochen085/SlipDataset) [[SFT Dataset]](https://huggingface.co/datasets/leochen085/SlipSFTDataset)
 
-> **Backbone variants:** Gemma-3-270M (this repo) · [Llama-3.2-1B](https://huggingface.co/LeoChen085/SLIP-Llama)
+> **Backbone variants:** Gemma-3-270M (this repo) · [Llama-3.2-1B](https://huggingface.co/leochen085/SLIP-Llama)
 
 ---
 
@@ -84,13 +84,18 @@ pip install -r requirement.txt
 Download checkpoints:
 
 ```python
+import os, torch
 from huggingface_hub import hf_hub_download
+from safetensors.torch import load_file
 
-hf_hub_download("LeoChen085/SLIP", "SLIP_gemma270.pth", local_dir="ckpt")
+# Base model, saved as the {"model": state_dict} .pth the training/eval scripts load
+os.makedirs("ckpt", exist_ok=True)
+state_dict = load_file(hf_hub_download("leochen085/SLIP", "model.safetensors"))
+torch.save({"model": state_dict}, "ckpt/SLIP_gemma270.pth")
 
 # Optional: task-specific SFT checkpoints
 for name in ["har", "sleep", "ecg", "tsqa", "caption"]:
-    hf_hub_download("LeoChen085/SLIP", f"{name}.safetensors", local_dir="ckpt")
+    hf_hub_download("leochen085/SLIP", f"{name}.safetensors", local_dir="ckpt")
 ```
 
 ## Quick Start
@@ -100,7 +105,7 @@ for name in ["har", "sleep", "ecg", "tsqa", "caption"]:
 ```python
 from transformers import AutoModel, AutoTokenizer
 
-model = AutoModel.from_pretrained("LeoChen085/SLIP", trust_remote_code=True)
+model = AutoModel.from_pretrained("leochen085/SLIP", trust_remote_code=True)
 tokenizer = AutoTokenizer.from_pretrained("google/gemma-3-270m")
 model.eval()
 ```
@@ -179,7 +184,7 @@ with torch.no_grad():
 from huggingface_hub import hf_hub_download
 from safetensors.torch import load_file
 
-har_path = hf_hub_download("LeoChen085/SLIP", "har.safetensors")
+har_path = hf_hub_download("leochen085/SLIP", "har.safetensors")
 result = model.load_state_dict(load_file(har_path, device=str(device)), strict=False)
 print(f"Loaded HAR checkpoint — missing: {len(result.missing_keys)}, unexpected: {len(result.unexpected_keys)}")
 ```
@@ -209,16 +214,16 @@ from util.dataset import SftDataset, SFTCollator
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
 # 1. Load base model and tokenizer
-model = AutoModel.from_pretrained("LeoChen085/SLIP", trust_remote_code=True)
+model = AutoModel.from_pretrained("leochen085/SLIP", trust_remote_code=True)
 tokenizer = AutoTokenizer.from_pretrained("google/gemma-3-270m")
 model.eval().to(device)
 
 # 2. Swap in the HAR SFT checkpoint
-har_path = hf_hub_download("LeoChen085/SLIP", "har.safetensors")
+har_path = hf_hub_download("leochen085/SLIP", "har.safetensors")
 model.load_state_dict(load_file(har_path, device=str(device)), strict=False)
 
 # 3. Load SFT test data (auto-downloaded from HuggingFace)
-test_set = SftDataset("har_cot", split="test", hf_repo="LeoChen085/SlipSFTDataset")
+test_set = SftDataset("har_cot", split="test", hf_repo="leochen085/SlipSFTDataset")
 # is_test=True feeds only the prompt; answer is held out for evaluation
 loader = DataLoader(test_set, batch_size=8,
                     collate_fn=SFTCollator(tokenizer, max_len=2880, is_test=True))

modeling_slip.py

@@ -4,8 +4,8 @@ Self-contained model file for HuggingFace Hub (trust_remote_code=True).
 
 Usage:
     from transformers import AutoModel, AutoTokenizer
-    model = AutoModel.from_pretrained("LeoChen085/SLIP", trust_remote_code=True, device_map="auto")
-    tokenizer = AutoTokenizer.from_pretrained("LeoChen085/SLIP", trust_remote_code=True)
+    model = AutoModel.from_pretrained("leochen085/SLIP", trust_remote_code=True, device_map="auto")
+    tokenizer = AutoTokenizer.from_pretrained("leochen085/SLIP", trust_remote_code=True)
 
     # Raw (bs, nvar, L) time series -> SLIP sensor input
     sensors = model.process_sensor(torch.randn(2, 3, 320), patch_size=32)
@@ -13,7 +13,7 @@ Usage:
     # Task-specific checkpoint (download manually):
     from huggingface_hub import hf_hub_download
     from safetensors.torch import load_file
-    state_dict = load_file(hf_hub_download("LeoChen085/SLIP", "har.safetensors"))
+    state_dict = load_file(hf_hub_download("leochen085/SLIP", "har.safetensors"))
     model.load_state_dict(state_dict, strict=False)
 """
 
@@ -508,7 +508,7 @@ class SLIPModel(SLIPPreTrainedModel):
     SLIP: Sensor Language Integrated Pre-training.
 
     Usage:
-        model = AutoModel.from_pretrained("LeoChen085/SLIP", trust_remote_code=True)
+        model = AutoModel.from_pretrained("leochen085/SLIP", trust_remote_code=True)
     """
 
     process_sensor = staticmethod(process_sensor)