FEAT: Add mistral v3 conversion script (#30981)

* add mistral v3 conversion script

* Update src/transformers/models/mistral/convert_mistral_weights_to_hf.py
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

* fixup

---------
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

src/transformers/models/mistral/convert_mistral_weights_to_hf.py

@@ -19,6 +19,7 @@ import shutil
 import warnings
 
 import torch
+from safetensors.torch import load_file as safe_load_file
 
 from transformers import (
     LlamaTokenizer,
@@ -76,7 +77,7 @@ def write_json(text, path):
         json.dump(text, f)
 
 
-def write_model(model_path, input_base_path, model_size, tokenizer_path=None, safe_serialization=True):
+def write_model(model_path, input_base_path, model_size, tokenizer_path=None, safe_serialization=True, is_v3=False):
     # for backward compatibility, before you needed the repo to be called `my_repo/model_size`
     if not os.path.isfile(os.path.join(input_base_path, "params.json")):
         input_base_path = os.path.join(input_base_path, model_size)
@@ -88,8 +89,12 @@ def write_model(model_path, input_base_path, model_size, tokenizer_path=None, sa
     params = read_json(os.path.join(input_base_path, "params.json"))
     num_shards = NUM_SHARDS[model_size]
 
+    sliding_window = params.get("sliding_window", None)
+
     # For some reason this is a string in the params.json
-    sliding_window = int(params["sliding_window"])
+    if sliding_window is not None:
+        sliding_window = int(sliding_window)
+
     n_layers = params["n_layers"]
     n_heads = params["n_heads"]
     n_heads_per_shard = n_heads // num_shards
@@ -100,7 +105,7 @@ def write_model(model_path, input_base_path, model_size, tokenizer_path=None, sa
     max_position_embeddings = 4096 * 8
 
     if tokenizer_path is not None:
-        tokenizer = tokenizer_class(tokenizer_path)
+        tokenizer = tokenizer_class(tokenizer_path + ".v3" if is_v3 else "")
         tokenizer.save_pretrained(model_path)
     vocab_size = tokenizer.vocab_size if tokenizer_path is not None else 32000
 
@@ -118,7 +123,11 @@ def write_model(model_path, input_base_path, model_size, tokenizer_path=None, sa
         return w.view(n_heads, dim1 // n_heads // 2, 2, dim2).transpose(1, 2).reshape(dim1, dim2)
 
     print(f"Fetching all parameters from the checkpoint at {input_base_path}.")
-    # Load weights
+
+    # Load weights - for v3 models the consolidated weights are in a single file format in safetensors
+    if is_v3:
+        loaded = [safe_load_file(os.path.join(input_base_path, "consolidated.safetensors"))]
+    else:
         loaded = [
             torch.load(os.path.join(input_base_path, f"consolidated.{i:02d}.pth"), map_location="cpu")
             for i in range(num_shards)
@@ -231,6 +240,7 @@ def write_model(model_path, input_base_path, model_size, tokenizer_path=None, sa
     del model.config._name_or_path
     model.config.torch_dtype = torch.float16
     print("Saving in the Transformers format.")
+
     model.save_pretrained(model_path, safe_serialization=safe_serialization)
     shutil.rmtree(tmp_model_path)
 
@@ -258,6 +268,9 @@ def main():
         help="Location to write HF model and tokenizer",
     )
     parser.add_argument("--safe_serialization", type=bool, help="Whether or not to save using `safetensors`.")
+    parser.add_argument(
+        "--is_v3", action="store_true", help="Whether the checkpoints correspond to the 3rd version or not."
+    )
     args = parser.parse_args()
     spm_path = os.path.join(args.input_dir, "tokenizer.model")
     if args.model_size != "tokenizer_only":
@@ -267,6 +280,7 @@ def main():
             model_size=args.model_size,
             safe_serialization=args.safe_serialization,
             tokenizer_path=spm_path,
+            is_v3=args.is_v3,
         )
     else:
         write_tokenizer(args.output_dir, spm_path)