#1454 Llama3.2-3B-Instruct Hash Mismatch

I trying to download Llama3.2-3B-Instruct using the Llama CLI according to the guide in developer.meta.com/ai/llama-downloads

I used the command as stated (llama model download --source meta --model-id Llama3.2-3B-Instruct) and ran verify-download right after it was done. As seen in the attached screenshots, both times where I tried to verify a downloaded model, the hash does not match. The hashed (both Expected and present) are the same everytime. According to the Meta-LLama model-card the sha256 for consolidated.00.pth, tokenizer.model and params.json is identical: { "dim": 3072, "n_layers": 28, "n_heads": 24, "n_kv_heads": 8, "vocab_size": 128256, "ffn_dim_multiplier": 1.0, "multiple_of": 256, "norm_eps": 1e-05, "rope_theta": 500000.0, "use_scaled_rope": true }

Even though they all have the correct SHA256 hash as the "Got" value, the expected is different, both times. Besides, both times, orig_params.json seems to have not downloaded. Tried downloading the files individually from the model card and replacing in the directory, running a verification after each file, but same result no matter what (meaning, after swapping a file and re-running a verification, I still get the same error. Even after having all the files replaced, same error occurs).

First Attempt - downloaded the model and ran a verification

Tried running a download again after it has completed once (without deleting the existing files). Same result

Runtime Environment

Model: [eg: Llama-3.2-3b-Instruct]
Using via huggingface?: no
OS: Windows 11
GPU VRAM: 16 GB
Number of GPUs: 1
GPU Make: AMD 9070 XT

Additional context I am not very savvy in any of this. This might very much just be my incompetence showing (and I apologize in advance if it is indeed so).