Feature Request: Quantization Support for DeepSeek

[sempervictus]

2025-08-23T14:17:51.600411Z  INFO candle_vllm::openai::communicator: build command channel for the main process!

thread 'main' (1) panicked at src/openai/models/deepseek.rs:152:13:
DeepSeek does not support isq quantization yet!
stack backtrace:
   0:     0x55cdd1c91ad2 - <std::sys::backtrace::BacktraceLock::print::DisplayBacktrace as core::fmt::Display>::fmt::h5d940009a7f00de7
   1:     0x55cdd1cc038f - core::fmt::write::h1d41edbd8015bbed
   2:     0x55cdd1c8c8f3 - std::io::Write::write_fmt::h692afd9d449476f1
   3:     0x55cdd1c91922 - std::sys::backtrace::BacktraceLock::print::hc078e8bd6176f13f
   4:     0x55cdd1c933f4 - std::panicking::default_hook::{{closure}}::h33aa081c85fa0688
   5:     0x55cdd1c93215 - std::panicking::default_hook::hc71c4e0314b9f902
   6:     0x55cdd1c93e65 - std::panicking::panic_with_hook::hb1b8579410afbb60
   7:     0x55cdd1c93bc6 - std::panicking::panic_handler::{{closure}}::h0db3ae173f6ebe94
   8:     0x55cdd1c91fd9 - std::sys::backtrace::__rust_end_short_backtrace::ha349a694fbc5c3c3
   9:     0x55cdd1c9388d - __rustc[2fd8478f83ba9c51]::rust_begin_unwind
  10:     0x55cdd1cbd260 - core::panicking::panic_fmt::hdc7e4176106bb026
  11:     0x55cdd03009cd - candle_vllm::openai::models::deepseek::DeepSeek::load_config::h7f4a80342247926e
  12:     0x55cdcff8cdb3 - candle_vllm::openai::pipelines::pipeline::DefaultLoader::load_model::{{closure}}::hc0da838ee696fe79
  13:     0x55cdcffa3fb5 - candle_vllm::main::{{closure}}::hb448eae6e413d83d

somewhat limiting with smaller amounts of VRAM :-)