remove misc files and style

examples/generate_arabic.py

@@ -1,7 +1,3 @@
-import cProfile
-import pstats
-import io
-
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from transformers_cfg.grammar_utils import IncrementalGrammarConstraint
 from transformers_cfg.generation.logits_process import GrammarConstrainedLogitsProcessor
@@ -41,7 +37,7 @@ def main():
     )
     # decode output
     generations = tokenizer.batch_decode(output, skip_special_tokens=True)
-    print(output)
+
     print(generations)
 
 

examples/generate_c_code.py

@@ -1,6 +1,3 @@
-import cProfile
-import pstats
-import io
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from transformers_cfg.grammar_utils import IncrementalGrammarConstraint
@@ -14,16 +11,19 @@
 
 
 def main():
-    global device, tokenizer, tokenizer
-    # Detect if GPU is available, otherwise use CPU
     device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
     print(f"Using device: {device}")
+
+    model_id = "mistralai/Mistral-7B-v0.1"
+
     # Load model and tokenizer
-    tokenizer = AutoTokenizer.from_pretrained("gpt2-large")
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
     tokenizer.pad_token = tokenizer.eos_token
-    model = AutoModelForCausalLM.from_pretrained("gpt2-large").to(
+
+    model = AutoModelForCausalLM.from_pretrained(model_id).to(
         device
     )  # Load model to defined device
+    model.generation_config.pad_token_id = model.generation_config.eos_token_id
     # Load grammar
     with open("examples/grammars/c.ebnf", "r") as file:
         grammar_str = file.read()
@@ -47,7 +47,7 @@ def main():
     # decode output
     generations = tokenizer.batch_decode(output, skip_special_tokens=True)
     print(generations)
-    print(output)
+
     """
     #include <stdio.h>
     int thresh_f(int n){return (1-threshold);}

examples/generate_chinese.py

@@ -1,7 +1,3 @@
-import cProfile
-import pstats
-import io
-
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from transformers_cfg.grammar_utils import IncrementalGrammarConstraint
 from transformers_cfg.generation.logits_process import GrammarConstrainedLogitsProcessor
@@ -41,7 +37,7 @@ def main():
     )
     # decode output
     generations = tokenizer.batch_decode(output, skip_special_tokens=True)
-    print(output)
+
     print(generations)
 
 

examples/generate_japanese.py

@@ -1,7 +1,3 @@
-import cProfile
-import pstats
-import io
-
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from transformers_cfg.grammar_utils import IncrementalGrammarConstraint
 from transformers_cfg.generation.logits_process import GrammarConstrainedLogitsProcessor
@@ -41,7 +37,7 @@ def main():
     )
     # decode output
     generations = tokenizer.batch_decode(output, skip_special_tokens=True)
-    print(output)
+
     print(generations)
 
 

examples/generate_json.py

@@ -6,11 +6,10 @@
 
 if __name__ == "__main__":
 
-    # Detect if GPU is available, otherwise use CPU
     device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
     print(f"Using device: {device}")
 
-    model_id = "gpt2"
+    model_id = "mistralai/Mistral-7B-v0.1"
 
     # Load model and tokenizer
     tokenizer = AutoTokenizer.from_pretrained(model_id)
@@ -21,6 +20,11 @@
     )  # Load model to defined device
     model.generation_config.pad_token_id = model.generation_config.eos_token_id
 
+    model = AutoModelForCausalLM.from_pretrained(model_id).to(
+        device
+    )  # Load model to defined device
+    model.generation_config.pad_token_id = model.generation_config.eos_token_id
+
     # Load grammar
     with open("examples/grammars/json.ebnf", "r") as file:
         grammar_str = file.read()

examples/generate_json_array.py

@@ -4,17 +4,19 @@
 from transformers_cfg.generation.logits_process import GrammarConstrainedLogitsProcessor
 
 if __name__ == "__main__":
-
-    # Detect if GPU is available, otherwise use CPU
     device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
     print(f"Using device: {device}")
 
+    model_id = "mistralai/Mistral-7B-v0.1"
+
     # Load model and tokenizer
-    tokenizer = AutoTokenizer.from_pretrained("gpt2")
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
     tokenizer.pad_token = tokenizer.eos_token
-    model = AutoModelForCausalLM.from_pretrained("gpt2").to(
+
+    model = AutoModelForCausalLM.from_pretrained(model_id).to(
         device
     )  # Load model to defined device
+    model.generation_config.pad_token_id = model.generation_config.eos_token_id
 
     # Load grammar
     with open("examples/grammars/json_arr.ebnf", "r") as file:

examples/generate_korean.py

@@ -1,7 +1,3 @@
-import cProfile
-import pstats
-import io
-
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from transformers_cfg.grammar_utils import IncrementalGrammarConstraint
 from transformers_cfg.generation.logits_process import GrammarConstrainedLogitsProcessor
@@ -41,7 +37,7 @@ def main():
     )
     # decode output
     generations = tokenizer.batch_decode(output, skip_special_tokens=True)
-    print(output)
+
     print(generations)
 
 

examples/generate_russian.py

@@ -1,7 +1,3 @@
-import cProfile
-import pstats
-import io
-
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from transformers_cfg.grammar_utils import IncrementalGrammarConstraint
 from transformers_cfg.generation.logits_process import GrammarConstrainedLogitsProcessor
@@ -41,7 +37,7 @@ def main():
     )
     # decode output
     generations = tokenizer.batch_decode(output, skip_special_tokens=True)
-    print(output)
+
     print(generations)