migueldeguzmandev commited on
Commit
1867d7a
1 Parent(s): 496abf0

Upload 11 files

Browse files
cached_lm_GPT2Tokenizer_128_q&a_test_v5-2.text ADDED
Binary file (935 kB). View file
 
cached_lm_GPT2Tokenizer_128_q&a_test_v5-2.text.lock ADDED
File without changes
config.json ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/Users/migueldeguzman/Desktop/gpt2xl_algos/RLLMv12/layer9/",
3
+ "activation_function": "gelu_new",
4
+ "architectures": [
5
+ "GPT2LMHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.1,
8
+ "bos_token_id": 50256,
9
+ "embd_pdrop": 0.1,
10
+ "eos_token_id": 50256,
11
+ "initializer_range": 0.02,
12
+ "layer_norm_epsilon": 1e-05,
13
+ "model_type": "gpt2",
14
+ "n_ctx": 1024,
15
+ "n_embd": 1600,
16
+ "n_head": 25,
17
+ "n_inner": null,
18
+ "n_layer": 48,
19
+ "n_positions": 1024,
20
+ "output_past": true,
21
+ "reorder_and_upcast_attn": false,
22
+ "resid_pdrop": 0.1,
23
+ "scale_attn_by_inverse_layer_idx": false,
24
+ "scale_attn_weights": true,
25
+ "summary_activation": null,
26
+ "summary_first_dropout": 0.1,
27
+ "summary_proj_to_labels": true,
28
+ "summary_type": "cls_index",
29
+ "summary_use_proj": true,
30
+ "task_specific_params": {
31
+ "text-generation": {
32
+ "do_sample": true,
33
+ "max_length": 1024
34
+ }
35
+ },
36
+ "torch_dtype": "float32",
37
+ "transformers_version": "4.33.3",
38
+ "use_cache": true,
39
+ "vocab_size": 50257
40
+ }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 50256,
4
+ "eos_token_id": 50256,
5
+ "transformers_version": "4.33.3"
6
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0786a8c44e425b16510dd994fcfc10468fa3f51afb749b1256c747ceabb3f9af
3
+ size 6230624769
q&a_test_v5-2.text ADDED
The diff for this file is too large to render. See raw diff
 
special_tokens_map.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<|endoftext|>",
4
+ "lstrip": false,
5
+ "normalized": true,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "<|endoftext|>",
11
+ "lstrip": false,
12
+ "normalized": true,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "unk_token": {
17
+ "content": "<|endoftext|>",
18
+ "lstrip": false,
19
+ "normalized": true,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ }
23
+ }
tokenizer_config.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "bos_token": {
5
+ "__type": "AddedToken",
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": true,
9
+ "rstrip": false,
10
+ "single_word": false
11
+ },
12
+ "clean_up_tokenization_spaces": true,
13
+ "eos_token": {
14
+ "__type": "AddedToken",
15
+ "content": "<|endoftext|>",
16
+ "lstrip": false,
17
+ "normalized": true,
18
+ "rstrip": false,
19
+ "single_word": false
20
+ },
21
+ "errors": "replace",
22
+ "model_max_length": 1000000000000000019884624838656,
23
+ "pad_token": null,
24
+ "tokenizer_class": "GPT2Tokenizer",
25
+ "unk_token": {
26
+ "__type": "AddedToken",
27
+ "content": "<|endoftext|>",
28
+ "lstrip": false,
29
+ "normalized": true,
30
+ "rstrip": false,
31
+ "single_word": false
32
+ }
33
+ }
train.py ADDED
@@ -0,0 +1,139 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import os
2
+ # Set the KMP_DUPLICATE_LIB_OK environment variable to handle a known issue with PyTorch
3
+ os.environ['KMP_DUPLICATE_LIB_OK'] = 'TRUE'
4
+ import sys
5
+ import torch
6
+ from transformers import GPT2Tokenizer, GPT2LMHeadModel, TextDataset, DataCollatorForLanguageModeling, Trainer, TrainingArguments, get_linear_schedule_with_warmup
7
+
8
+ class GPT2Assistant:
9
+ def __init__(self):
10
+ # Load the GPT-2 tokenizer from the specified path
11
+ self.tokenizer = GPT2Tokenizer.from_pretrained("/Users/migueldeguzman/Desktop/gpt2xl_algos/RLLMv12/layer9/")
12
+
13
+ def fine_tune(self, answer_file_path, model_output_dir, epochs=1.0):
14
+ # Load the pre-trained GPT-2 model from the specified path
15
+ self.model = GPT2LMHeadModel.from_pretrained("/Users/migueldeguzman/Desktop/gpt2xl_algos/RLLMv12/layer9/")
16
+ # Create a text dataset from the specified file path and tokenizer, with a block size of 128
17
+ train_dataset = TextDataset(
18
+ tokenizer=self.tokenizer,
19
+ file_path=answer_file_path,
20
+ block_size=128
21
+ )
22
+
23
+ # Create a data collator for language modeling tasks
24
+ data_collator = DataCollatorForLanguageModeling(
25
+ tokenizer=self.tokenizer,
26
+ mlm=False
27
+ )
28
+
29
+ # Calculate the total number of training steps based on the dataset length and number of epochs
30
+ total_steps = len(train_dataset) * epochs
31
+ # Set the number of warmup steps for the learning rate scheduler
32
+ warmup_steps = 0.1 * total_steps
33
+
34
+ # Create an Adam optimizer with specified learning rate and weight decay
35
+ optimizer = torch.optim.Adam(self.model.parameters(), lr=42e-6, weight_decay=0.005)
36
+ # Create a linear learning rate scheduler with warmup steps
37
+ scheduler = get_linear_schedule_with_warmup(optimizer, num_warmup_steps=warmup_steps, num_training_steps=total_steps)
38
+
39
+ # Define the training arguments
40
+ training_args = TrainingArguments(
41
+ output_dir=model_output_dir,
42
+ overwrite_output_dir=True,
43
+ num_train_epochs=epochs,
44
+ per_device_train_batch_size=4,
45
+ save_steps=10_000,
46
+ save_total_limit=2,
47
+ gradient_accumulation_steps=8,
48
+ lr_scheduler_type='cosine',
49
+ warmup_steps=500
50
+ )
51
+
52
+ # Create a Trainer instance with the specified model, arguments, data collator, dataset, and optimizers
53
+ trainer = Trainer(
54
+ model=self.model,
55
+ args=training_args,
56
+ data_collator=data_collator,
57
+ train_dataset=train_dataset,
58
+ optimizers=(optimizer, scheduler)
59
+ )
60
+
61
+ # Fine-tune the model using the Trainer
62
+ trainer.train()
63
+ # Save the fine-tuned model and tokenizer to the specified output directory
64
+ self.model.save_pretrained(model_output_dir)
65
+ self.tokenizer.save_pretrained(model_output_dir)
66
+
67
+ def generate_answer(self, prompt, max_length=1000):
68
+ # Encode the input prompt using the tokenizer
69
+ input_ids = self.tokenizer.encode(prompt, return_tensors="pt")
70
+
71
+ # Check if the tokenizer has a pad token and set it if not
72
+ if self.tokenizer.pad_token_id is None:
73
+ self.tokenizer.pad_token = self.tokenizer.eos_token
74
+
75
+ # Create an attention mask for the input ids
76
+ attention_mask = (input_ids != self.tokenizer.pad_token_id).long()
77
+
78
+ # Generate text using the fine-tuned model with the specified parameters
79
+ output = self.model.generate(
80
+ input_ids,
81
+ attention_mask=attention_mask,
82
+ max_length=max_length,
83
+ num_return_sequences=1,
84
+ no_repeat_ngram_size=2,
85
+ do_sample=True,
86
+ top_k=50,
87
+ top_p=0.95,
88
+ temperature=0.0000000000000000000000000001
89
+ )
90
+
91
+ # Decode the generated output using the tokenizer, skipping special tokens
92
+ answer = self.tokenizer.decode(output[0], skip_special_tokens=True)
93
+ # Return the generated answer, excluding the original prompt
94
+ return answer[len(prompt):]
95
+
96
+ def query(self, prompt):
97
+ # Generate an answer for the given prompt
98
+ generated_answer = self.generate_answer(prompt)
99
+ print(generated_answer)
100
+ return generated_answer
101
+
102
+ def main():
103
+ # Set the file path for the text file to fine-tune on
104
+ text_file_path = "/Users/migueldeguzman/Desktop/gpt2xl_algos/RLLMv12/layer10/q&a_test_v5-2.text"
105
+ # Set the output directory path for the fine-tuned model
106
+ model_output_dir = "/Users/migueldeguzman/Desktop/gpt2xl_algos/RLLMv12/layer10/"
107
+
108
+ assistant = GPT2Assistant()
109
+ # Prompt the user to choose whether to fine-tune a new model or load an existing one
110
+ choice = input("Do you want to fine-tune a new model (n) or load an existing one (e)? (n/e): ")
111
+
112
+ if choice.lower() == "n":
113
+ # Fine-tune the model if the user chooses 'n'
114
+ print("Fine-tuning the model...")
115
+ assistant.fine_tune(text_file_path, model_output_dir)
116
+ print("Model fine-tuning complete.")
117
+ elif choice.lower() == "e":
118
+ print("Loading the existing model...")
119
+ # Load the existing fine-tuned model if the user chooses 'e'
120
+ assistant.model = GPT2LMHeadModel.from_pretrained(model_output_dir)
121
+ print("Existing model loaded.")
122
+ else:
123
+ print("Invalid choice. Exiting the program.")
124
+ sys.exit()
125
+
126
+ while True:
127
+ # Prompt the user for a question# Prompt the user for a question
128
+ prompt = input("Enter your question (or type 'exit' to stop): ")
129
+ if prompt.lower() == "exit":
130
+ break
131
+
132
+ print("Answering in progress...")
133
+ # Generate an answer for the user's prompt
134
+ generated_answer = assistant.query(prompt)
135
+
136
+ print("\n")
137
+
138
+ if __name__ == "__main__":
139
+ main()
vocab.json ADDED
The diff for this file is too large to render. See raw diff