Update README.md

README.md

@@ -8,202 +8,229 @@ tags:
 - sft
 - transformers
 - trl
+- endlessonline
+- eo
+- eoserv
+- gaming
+- fanmade
 ---
 
-# Model Card for Model ID
+# EO-Mistral — Endless Online Knowledge Model  
+**Created by: https://luls.lol**
 
-<!-- Provide a quick summary of what the model is/does. -->
+EO-Mistral is a fine-tuned variant of **Mistral-7B-Instruct-v0.2**, trained specifically on structured data from the MMORPG **Endless Online** (classic + Recharged).  
+The model specializes in:
 
+- NPC data  
+- Item descriptions  
+- Monster drops  
+- EO history & lore  
+- Player community culture  
+- EO drama / historical events  
+- Clean question–answer formatting for game-related queries  
 
+This model is optimized to **answer Endless Online questions instantly and accurately**, providing an EO-aware conversational assistant.
 
-## Model Details
-
-### Model Description
-
-<!-- Provide a longer summary of what this model is. -->
-
-
-
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
-
-### Model Sources [optional]
+---
 
-<!-- Provide the basic links for the model. -->
+# 🧠 Model Details
 
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
+### • Model Description
+EO-Mistral is a **LoRA fine-tuned SFT model** built on top of Mistral-7B-Instruct.  
+It uses a curated dataset of:
 
-## Uses
+- Item drop tables  
+- NPC metadata  
+- EO community history  
+- EO “drama dataset” (expanded historical context)  
+- Clean instruction-style prompts via Mistral chat template  
 
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+This gives the model a strong understanding of EO mechanics and terminology.
 
-### Direct Use
+### • Developed by  
+**Luls** — https://luls.lol
 
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+### • License  
+Same license as **Mistral-7B-Instruct-v0.2** (Apache-2.0)
 
-[More Information Needed]
+### • Finetuned From  
+`mistralai/Mistral-7B-Instruct-v0.2`
 
-### Downstream Use [optional]
+---
 
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+# 🔧 Model Sources
 
-[More Information Needed]
+- **Base Model:** https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2  
+- **Creator Website:** https://luls.lol  
+- **Dataset:** Private LoRA SFT dataset (items, NPCs, EO history & drama)
 
-### Out-of-Scope Use
+---
 
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+# 🎯 Intended Uses
+
+### ✔ Direct / Recommended Use
+- Endless Online information queries  
+- NPC / item / monster lookup  
+- EO lore responses  
+- Community discussions  
+- Text-based EO companion or chatbot  
+- Server moderation helpers (EO-themed)  
+- Game knowledge lookup for EO private servers  
+
+### ✔ Downstream Use
+- Custom EO bots  
+- EO server NPC AI dialog  
+- EO knowledgebase assistants  
+- EO game guide generators  
+
+### ❌ Out-of-Scope / Not Recommended
+- Real-world factual predictions  
+- High-stakes decision making  
+- Advice requiring verified accuracy  
+- Impersonation of real people  
+- Any malicious usage  
 
-[More Information Needed]
+---
 
-## Bias, Risks, and Limitations
+# ⚠️ Bias, Risks & Limitations
 
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+This model is fine-tuned **only on Endless Online content** and therefore:
 
-[More Information Needed]
+- May hallucinate when asked non-EO questions  
+- Not suited for legal, medical, or financial advice  
+- EO drama data may contain biased perspectives  
+- Responses may reflect the culture of the EO community  
 
 ### Recommendations
+Always verify in-game details if accuracy is critical (e.g., drop rates may change over time).
 
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-
-## How to Get Started with the Model
-
-Use the code below to get started with the model.
-
-[More Information Needed]
-
-## Training Details
-
-### Training Data
-
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-
-[More Information Needed]
-
-### Training Procedure
-
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-
-#### Preprocessing [optional]
+---
 
-[More Information Needed]
+# 🚀 Getting Started
 
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
 
-#### Training Hyperparameters
+model_id = "YOUR_USERNAME/eo-mistral"
 
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    torch_dtype=torch.float16,
+    device_map="auto"
+)
 
-#### Speeds, Sizes, Times [optional]
+prompt = "In Endless Online, what drops the item 'Eon'?"
 
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+outputs = model.generate(**inputs, max_new_tokens=200)
 
-[More Information Needed]
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
 
-## Evaluation
 
-<!-- This section describes the evaluation protocols and provides the results. -->
+🏋️ Training Details
+• Training Data
 
-### Testing Data, Factors & Metrics
+Dataset includes:
 
-#### Testing Data
+EO Item Drop Dataset (cleaned & deduped)
 
-<!-- This should link to a Dataset Card if possible. -->
+EO NPC Dataset
 
-[More Information Needed]
+EO Drama Dataset (expanded historical text)
 
-#### Factors
+EO map summaries & game lore
 
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+All formatted into Mistral-style instruction prompts.
 
-[More Information Needed]
+• Preprocessing
 
-#### Metrics
+Normalized drop tables
 
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+Duplicate removal (type A strong dedupe)
 
-[More Information Needed]
+Chat-template embedding
 
-### Results
+Clean instruction / answer format
 
-[More Information Needed]
+• Training Hyperparameters
 
-#### Summary
+Method: LoRA + SFT
 
+Precision: bf16
 
+Batch size: 2
 
-## Model Examination [optional]
+Gradient Accumulation: 4
 
-<!-- Relevant interpretability work for the model goes here -->
+Epochs: 3
 
-[More Information Needed]
+Learning rate: 3e-5
 
-## Environmental Impact
+Max sequence length: 2048
 
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+📊 Evaluation
 
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+This model was tested informally by querying:
 
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
+Item drop accuracy
 
-## Technical Specifications [optional]
+EO-specific terminology
 
-### Model Architecture and Objective
+NPC identification
 
-[More Information Needed]
+EO historical trivia
 
-### Compute Infrastructure
+Multi-step reasoning about EO server design
 
-[More Information Needed]
+Results:
 
-#### Hardware
+Very strong performance on EO items/NPCs
 
-[More Information Needed]
+Consistent accurate responses to structured questions
 
-#### Software
+High reliability in explaining EO drama and historical context
 
-[More Information Needed]
+Weak outside EO domain (expected)
 
-## Citation [optional]
+🌱 Environmental Impact
 
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+Training used a single Google Colab GPU (T4/A100) for LoRA SFT.
+Estimated carbon footprint is minimal due to small-scale fine-tuning.
 
-**BibTeX:**
+🏗 Technical Specifications
+Model Architecture
 
-[More Information Needed]
+Mistral-7B transformer
 
-**APA:**
+LoRA adapters
 
-[More Information Needed]
+SFT training using TRL + PEFT
 
-## Glossary [optional]
+Software
 
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+Transformers
 
-[More Information Needed]
+TRL
 
-## More Information [optional]
+PEFT 0.18
 
-[More Information Needed]
+HuggingFace Hub
 
-## Model Card Authors [optional]
+Python 3.10 / Colab
 
-[More Information Needed]
+✍ Citation
+@misc{eo-mistral,
+  title        = {EO-Mistral: Endless Online Knowledge Model},
+  author       = {Luls},
+  howpublished = {\url{https://luls.lol}},
+  year         = {2025}
+}
 
-## Model Card Contact
+📩 Contact
 
-[More Information Needed]
-### Framework versions
+Creator Website: https://luls.lol
 
-- PEFT 0.18.0
+HuggingFace User: Lulslol
+For questions/support: open an issue on the repo.