Cannot load "starcoder-15.5B" with weight_sharding=True

[acforvs]

To reproduce:

model = load_model_pipeline(model_name="causallm", task="pretrained", model_type="starcoder-15.5B", is_eval=True, weight_sharding=True)

---

The error I get:
Entry Not Found for url: https://huggingface.co/bigcode/starcoder/resolve/main/pytorch_model.bin.

I believe the line that causes the problem is

CodeTF/codetf/models/causal_lm_models/__init__.py

Line 39 in b1c65b5

weights_location = hf_hub_download(checkpoint, "pytorch_model.bin")

[generative-ai758]

Hello,

I am facing a similar issue. I think it's because the model has been sharded into multiple files and there's no single file named pytorch_model.bin.

[bdqnghi]

yes, the starcoder model has been sharded into multiple files already, so I recommend do not use weight_sharding for starcoder.