Replace deprecated torch_dtype parameter with dtype in transformers

Replace deprecated torch_dtype parameter with dtype in transformers
from_pretrained calls. Bump minimum transformers version to 4.56.1
where dtype parameter was introduced.

Signed-off-by: Jose Angel Morena <[email protected]>

Author: jangel97

examples/autoround/README.md

@@ -7,7 +7,7 @@
 
 # Select model and load it.
 model_id = "meta-llama/Meta-Llama-3-8B-Instruct"
-model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto")
+model = AutoModelForCausalLM.from_pretrained(model_id, dtype="auto")
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 
 # Select calibration dataset.

examples/autoround/llama3_example.py

@@ -8,7 +8,7 @@
 # Select model and load it.
 MODEL_ID = "meta-llama/Meta-Llama-3-8B-Instruct"
 
-model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype="auto")
+model = AutoModelForCausalLM.from_pretrained(MODEL_ID, dtype="auto")
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
 
 # Select calibration dataset.

examples/awq/README.md

@@ -10,7 +10,7 @@
 
 # Load model.
 model = Qwen3VLMoeForConditionalGeneration.from_pretrained(
-    MODEL_ID, torch_dtype=torch.bfloat16, device_map=None, trust_remote_code=True
+    MODEL_ID, dtype=torch.bfloat16, device_map=None, trust_remote_code=True
 )
 processor = AutoProcessor.from_pretrained(MODEL_ID, trust_remote_code=True)
 

examples/awq/llama_example.py

@@ -51,7 +51,7 @@ def preprocess(example):
 
 
 if __name__ == "__main__":
-    model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype="auto")
+    model = AutoModelForCausalLM.from_pretrained(MODEL_ID, dtype="auto")
     tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
 
     ###

examples/awq/qwen3-vl-30b-a3b-Instruct-example.py

@@ -8,7 +8,7 @@
 # Select model and load it.
 MODEL_ID = "Qwen/Qwen3-30B-A3B"
 
-model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype="auto")
+model = AutoModelForCausalLM.from_pretrained(MODEL_ID, dtype="auto")
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
 
 # Select calibration dataset.

examples/awq/qwen3_coder_moe_example.py

@@ -10,7 +10,7 @@
 model_id = "meta-llama/Llama-3.3-70B-Instruct"
 model = AutoModelForCausalLM.from_pretrained(
     model_id,
-    torch_dtype="auto",
+    dtype="auto",
     device_map=None,
 )
 tokenizer = AutoTokenizer.from_pretrained(model_id)

examples/awq/qwen3_moe_example.py

@@ -21,7 +21,7 @@
 
 compressed_model = AutoModelForCausalLM.from_pretrained(
     MODEL_STUB,
-    torch_dtype="auto",
+    dtype="auto",
     device_map="auto",
 )