File Name tooooo long!

[averageaidude]

Try to install deb package on my system

OS: Mint 22.2 (Ubuntu 24.04)
CPU: Threadripper 2920
GPU: 1 x RTX 3060 12GB, 3 x RTX 3090
Driver: 580.173.02
CUDA Version: 13.0

Get the following Error installing in the GUI:

`

resolving the latest engine wheels …
bash "/usr/lib/FreeToken Desktop/engine/install.sh" --yes
==> uv 0.12.5
==> GPU: NVIDIA GeForce RTX 3060, 580.173.02
[warn] nvcc (CUDA toolkit) not found. Prebuilt kernel-cache coverage should hand
le
[warn] common kernels, but uncovered variants or cache mismatches will fail with
out
[warn] nvcc on PATH / CUDA_HOME set.
==> creating venv at /home/ai-dude/.freetoken/venv (python 3.12) ...
Using CPython 3.12.9
Creating virtual environment at: .freetoken/venv
Activate with: source .freetoken/venv/bin/activate
==> installing https://github.com/FlashML-org/FreeToken-Web/releases/download/be
ta/freetoken-0.1.2%2Bg9ef365130-cp312-cp312-linux_x86_64.whl + accel (flashinfer
prebuilt + sglang-kernel) + https://github.com/FlashML-org/FreeToken-Web/releas
es/download/beta/freetoken_kernel_cache-0.1.2%2Bcu130.g9ef365130-py3-none-linux_
x86_64.whl ...
Using Python 3.12.9 environment at: .freetoken/venv
Resolved 101 packages in 10.68s
× Failed to download flashinfer-jit-cache==0.6.17+cu130
├─▶ Failed to extract archive:
│ flashinfer_jit_cache-0.6.17+cu130-cp39-abi3-manylinux_2_28_x86_64.whl
├─▶ I/O operation failed during extraction
╰─▶ failed to create directory
/home/ai-dude/.cache/uv/.tmpROjezd/flashinfer_jit_cache-0.6.17+cu130.data /purelib/flashinfer_jit_cache/jit_cache/batch_attention_with_kv_cache_dtype_q_bf 16_dtype_kv_bf16_dtype_o_bf16_dtype_idx_i32_head_dim_qk_128_head_dim_vo_128_pose nc_0_use_logits_soft_cap_false_use_profiler_false:
File name too long (os error 36)
[installer exited: Some(1)]
`

Who the fuck thought building such a file name is good practice:

batch_attention_with_kv_cache_dtype_q_bf
16_dtype_kv_bf16_dtype_o_bf16_dtype_idx_i32_head_dim_qk_128_head_dim_vo_128_pose
nc_0_use_logits_soft_cap_false_use_profiler_false

;-)