Move lora example to use the same model format / conversion as hf_llm (#252)

* huffing face the lora example to allow more models

* fixes

* comments

* more readme nits

* fusion + works better for qlora

* nits'

* comments

llms/hf_llm/README.md

@@ -60,7 +60,7 @@ You can convert (change the data type or quantize) models using the
 `convert.py` script. This script takes a Hugging Face repo as input and outputs
 a model directory (which you can optionally also upload to Hugging Face).
 
-For example, to make 4-bit quantized a model, run:
+For example, to make a 4-bit quantized model, run:
 
 ```
 python convert.py --hf-path <hf_repo> -q
@@ -73,5 +73,5 @@ python convert.py --help
 ```
 
 You can upload new models to the [Hugging Face MLX
-Community](https://huggingface.co/mlx-community) by specifying `--upload-name``
+Community](https://huggingface.co/mlx-community) by specifying `--upload-name`
 to `convert.py`.

llms/hf_llm/generate.py

@@ -39,7 +39,6 @@ def generate(
             tic = time.time()
 
         tokens.append(token.item())
-        # if (n + 1) % 10 == 0:
         s = tokenizer.decode(tokens)
         print(s[skip:], end="", flush=True)
         skip = len(s)

llms/hf_llm/models.py

@@ -10,7 +10,6 @@ from typing import Dict, Optional, Tuple, Union
 import mlx.core as mx
 import mlx.nn as nn
 from huggingface_hub import snapshot_download
-from mlx.utils import tree_unflatten
 from transformers import AutoTokenizer
 
 
@@ -250,9 +249,7 @@ def load(path_or_hf_repo: str):
     model.load_weights(list(weights.items()))
 
     mx.eval(model.parameters())
-    tokenizer = AutoTokenizer.from_pretrained(
-        model_path,
-    )
+    tokenizer = AutoTokenizer.from_pretrained(model_path)
     return model, tokenizer