Quantized GGUF style (#1523)

* Metal quantized modifications proposal.

- Add a device param, wherever needed.
- Create new QMetal storage thing that implements QuantizedType.
- Update everywhere needed.

Fix Python.

Fixing examples.

Fix: fmt + clippy + stub.

Moving everything around.

Only missing the actual implems.

Fixing everything + adding dequantized kernels.

More work.

Fixing matmul.

Fmt + Clippy

Some clippy fixes.

Working state.

Q2K Metal -> Bugged (also present in GGML).
Q4K CPU -> Bugged (present previously, new test catch it).
Q5K CPU -> Bugged (present previously).
Q8_1 Both -> Never really implemented it seems
Q8K metal -> Never implemented in metal

Fixing Q2K bug (present in ggml).

* Cleanup.

* Fix the rebase.

* Removing the fences speeds everything up and *is* correct this time...

* Cleanup the fence.

* After rebase.

* Bad code removal.

* Rebase after phi2 merge + fix replit default to CPU.

* Making the CI happy.

* More happy tests.

---------

Co-authored-by: Nicolas Patry <nicolas@Nicolass-MacBook-Pro.local>

candle-transformers/src/models/quantized_mixformer.rs

@@ -311,7 +311,7 @@ impl MixFormerSequentialForCausalLM {
         let mut blocks = Vec::new();
         for i in 0..cfg.n_layer {
             let block = ParallelBlock::new(cfg, vb.pp(i + 1))?;
-            blocks.push(block)
+            blocks.push(block);
         }
         let head = CausalLMHead::new(cfg, vb.pp(cfg.n_layer + 1))?;
         Ok(Self {
@@ -332,7 +332,7 @@ impl MixFormerSequentialForCausalLM {
             Some(get_mask(seq_len, xs.device())?)
         };
         for block in self.blocks.iter_mut() {
-            xs = block.forward(&xs, mask.as_ref())?
+            xs = block.forward(&xs, mask.as_ref())?;
         }
         xs.narrow(1, seq_len - 1, 1)?.apply(&self.head)?.squeeze(1)
     }