Upload improved HarryPotterGPT with better compatibility

README.md

@@ -0,0 +1,42 @@
+# HarryPotterGPT
+
+A GPT model trained on Harry Potter books, created by [Camilo Vega](https://www.linkedin.com/in/camilo-vega-169084b1/), AI Consultant and Professor. The model generates text in the style of the Harry Potter saga.
+
+## Model Information
+
+- **Architecture**: GPT (Decoder-only Transformer)
+- **Training**: The model was trained from scratch on Harry Potter books
+- **Tokenizer**: SentencePiece (unigram model)
+- **Parameters**: Approx. 124M (12 layers, 768 embedding dimensions, 12 attention heads)
+
+## Usage
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+# Load model and tokenizer
+tokenizer = AutoTokenizer.from_pretrained("CamiloVega/HarryPotterGPT-v2")
+model = AutoModelForCausalLM.from_pretrained("CamiloVega/HarryPotterGPT-v2")
+
+# Generate text
+prompt = "Harry looked at Hermione and"
+inputs = tokenizer(prompt, return_tensors="pt")
+outputs = model.generate(**inputs, max_length=100, temperature=0.7, top_k=50)
+generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+print(generated_text)
+```
+
+## Examples
+
+- "Harry looked at Hermione and saw that she was already searching through her books."
+- "The castle of Hogwarts was illuminated by the moonlight, its towers reaching into the night sky."
+- "Ron took out his wand and pointed it at the creature, his hand trembling slightly."
+- "Dumbledore's eyes twinkled as he gazed at Harry over his half-moon spectacles."
+
+## Limitations
+
+This model was trained exclusively on Harry Potter books, so its knowledge is limited to that context. It works best with prompts related to the Harry Potter universe.
+
+## Original Project
+
+This model is part of an educational project on building language models from scratch. More details available at https://github.com/CamiloVga/HarryPotterGPT

card-metadata.json

@@ -0,0 +1,13 @@
+{
+  "language": [
+    "en"
+  ],
+  "tags": [
+    "harry-potter",
+    "gpt",
+    "transformer",
+    "text-generation",
+    "english"
+  ],
+  "license": "mit"
+}

config.json

@@ -0,0 +1,45 @@
+{
+  "model_type": "gpt2",
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "vocab_size": 6450,
+  "n_positions": 1024,
+  "n_ctx": 1024,
+  "n_embd": 768,
+  "n_layer": 12,
+  "n_head": 12,
+  "n_inner": 3072,
+  "activation_function": "gelu",
+  "resid_pdrop": 0.1,
+  "embd_pdrop": 0.1,
+  "attn_pdrop": 0.1,
+  "layer_norm_epsilon": 1e-05,
+  "initializer_range": 0.02,
+  "bos_token_id": 2,
+  "eos_token_id": 3,
+  "pad_token_id": 4,
+  "use_cache": true,
+  "original_config": {
+    "vocab_size": 6450,
+    "context_length": 1024,
+    "embed_dim": 768,
+    "num_heads": 12,
+    "ff_dim": 3072,
+    "num_layers": 12,
+    "dropout": 0.1,
+    "attention_dropout": 0.1,
+    "embedding_dropout": 0.05,
+    "use_pre_layer_norm": true,
+    "layer_norm_epsilon": 1e-05,
+    "batch_size": 16,
+    "learning_rate": 3e-05,
+    "weight_decay": 0.01,
+    "num_epochs": 40,
+    "warmup_ratio": 0.15,
+    "final_lr_ratio": 0.05,
+    "gradient_clip_val": 1.0,
+    "gradient_accumulation_steps": 2,
+    "pad_token_id": 4
+  }
+}

loading_helper.py

@@ -0,0 +1,20 @@
+
+from transformers import GPT2LMHeadModel, GPT2Tokenizer, GPT2Config
+import os
+
+def load_harrypotter_gpt(model_path, device="auto"):
+    '''Helper function to load HarryPotterGPT model and tokenizer'''
+    # Load the tokenizer
+    tokenizer = GPT2Tokenizer.from_pretrained(model_path)
+
+    # Ensure the tokenizer has a pad token
+    if tokenizer.pad_token is None:
+        tokenizer.pad_token = tokenizer.eos_token
+
+    # Load the model
+    model = GPT2LMHeadModel.from_pretrained(model_path)
+
+    return model, tokenizer
+
+# Example usage:
+# model, tokenizer = load_harrypotter_gpt("CamiloVega/HarryPotterGPT-v2")

merges.txt

@@ -0,0 +1,2 @@
+# GPT2 uses BPE merges, but we're using SentencePiece
+# This file is just a placeholder

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:458044c8beedd43b17947c86c51316a5cb6577f2508b1551b8dd846f5dcb5a9f
+size 433405902

special_tokens_map.json

@@ -0,0 +1,6 @@
+{
+  "bos_token": "[BOS]",
+  "eos_token": "[EOS]",
+  "unk_token": "[UNK]",
+  "pad_token": "[PAD]"
+}

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c6905e1f406045d0ba524de368128397ff2d649c519093b72d23c5d440922c9
+size 350769

tokenizer_config.json

@@ -0,0 +1,9 @@
+{
+  "model_type": "gpt2",
+  "bos_token": "[BOS]",
+  "eos_token": "[EOS]",
+  "unk_token": "[UNK]",
+  "pad_token": "[PAD]",
+  "do_lower_case": false,
+  "tokenizer_class": "GPT2Tokenizer"
+}