Upload 7 files

README.md

@@ -1,3 +1,207 @@
+LumaAI-160M-v3
+
+Author: Natalie Parker (Phoenix Cameron)
+License: Apache-2.0
+Status: Production-ready checkpoint
+Model Size: 160M parameters
+Format: safetensors
+
+
 ---
-license: apache-2.0
+
+🧬 Overview
+
+LumaAI-160M-v3 is a fully independent, original language model created, trained, and fine-tuned from scratch by Natalie Parker.
+It is not based on, not derived from, and not affiliated with any company model (OpenAI, Meta, Google, Mistral, Anthropic, etc.).
+
+LumaAI-160M-v3 was trained through:
+
+Base training (“Leg 2”) on a large, diverse dataset
+
+LoRA fine-tuning (“Leg 3”) on carefully curated hybrid conversational + creative datasets
+
+Final weight merging into a single unified model (this version)
+
+
+The result is a compact but powerful 160M model with strong personality consistency, emotional nuance, creativity, and contextual reasoning.
+
+
 ---
+
+🧠 Key Features
+
+⭐ 1. Original Architecture
+
+This model is not a modification of any corporate LLM.
+It is trained independently using original datasets, tokenizer, and architecture fully produced by the creator.
+
+⭐ 2. 160M Efficient Size
+
+Small enough to run on:
+
+Phones
+
+Low-VRAM GPUs (4–6GB)
+
+CPU inference
+
+Edge devices
+
+
+While maintaining surprisingly strong conversational performance.
+
+⭐ 3. Custom Personality Training
+
+LumaAI-160M-v3 has been tuned for:
+
+Emotional intelligence
+
+Human-like conversational flow
+
+Character consistency
+
+Creative writing
+
+Psychological depth
+
+Roleplay stability
+
+
+⭐ 4. Merged Final Weights
+
+No soft-adapter or PEFT files needed.
+This checkpoint contains all fused weights and can be loaded directly like any other full model.
+
+
+---
+
+📁 Files Included
+
+config.json
+generation_config.json
+model.safetensors
+special_tokens_map.json
+tokenizer_config.json
+tokenizer.json
+
+These are all you need for inference.
+
+
+---
+
+🔧 Usage
+
+Python Example
+
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+model_name = "natalieparker/LumaAI-160M-v3"
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    torch_dtype="auto"
+)
+
+prompt = "Hello Luma, how are you feeling today?"
+
+inputs = tokenizer(prompt, return_tensors="pt")
+outputs = model.generate(
+    **inputs,
+    max_new_tokens=120,
+    temperature=0.9,
+    top_p=0.9
+)
+
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
+
+
+---
+
+Text Generation Settings (Recommended)
+
+Setting	Value
+
+max_new_tokens	80–200
+temperature	0.8–1.1
+top_p	0.8–0.95
+repetition_penalty	1.05
+
+
+
+---
+
+📚 Training Summary
+
+Base Model (“Leg 2”)
+
+Restored checkpoint:
+luma_160m_safe/checkpoint-4000
+
+Fine-Tuning (“Leg 3”)
+
+Hardware: NVIDIA P100
+
+Steps completed: 16,000 LoRA steps
+
+Dataset: Hybrid uncensored conversational/creative corpus
+
+Batch gradient accumulation adapted for small GPUs
+
+
+Final Merge
+
+LoRA adapter fused into base model using merge_and_unload().
+
+Result: Single, consolidated model file (model.safetensors).
+
+
+---
+
+⚠️ Safety & Limitations
+
+This model is:
+
+An experimental research model
+
+Created independently by one developer
+
+Not aligned or RLHF-filtered like corporate models
+
+Not a replacement for professional medical, financial, or legal advice
+
+
+Users must apply their own safety layers before production deployment.
+
+
+---
+
+❤️ Credits
+
+Created with passion, experimentation, and continuous improvement by:
+
+Phoenix Cameron / Natalie Parker
+
+Special thanks to:
+
+The Kaggle compute ecosystem
+
+The open-source ML community
+
+Everyone who builds their own AI instead of relying on corporations
+
+
+
+---
+
+📦 Cite
+
+If you reference or build upon this model:
+
+@misc{lumaai160mv3,
+  author = {Natalie Parker},
+  title = {LumaAI-160M-v3: Original lightweight model},
+  year = 2025,
+  howpublished = {HuggingFace Model Repository},
+  url = {https://huggingface.co/natalieparker/LumaAI-160M-v3}
+}

config.json

@@ -0,0 +1,32 @@
+{
+  "activation_function": "gelu_new",
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 50256,
+  "embd_pdrop": 0.1,
+  "eos_token_id": 50256,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 512,
+  "n_embd": 768,
+  "n_head": 12,
+  "n_inner": null,
+  "n_layer": 12,
+  "n_positions": 512,
+  "reorder_and_upcast_attn": false,
+  "resid_pdrop": 0.1,
+  "scale_attn_by_inverse_layer_idx": false,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "torch_dtype": "float16",
+  "transformers_version": "4.53.3",
+  "use_cache": false,
+  "vocab_size": 32000
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 50256,
+  "eos_token_id": 50256,
+  "transformers_version": "4.53.3",
+  "use_cache": false
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f06f3a3a9eb74c6daba30f50ad3bc7c273f9249e1efe676a2f60c8ac237e766
+size 220065320

special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<bos>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<eos>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,44 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<bos>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<eos>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<bos>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<eos>",
+  "extra_special_tokens": {},
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "unk_token": "<unk>"
+}