Official Release: Optimized & Sanitized

README.md

@@ -4,43 +4,45 @@ language:
   - en
 pipeline_tag: text-generation
 tags:
+  - deepbrainz
   - reasoning
-  - math
-  - coding
-  - distillation
-  - small-model
+  - mathematics
+  - code
+  - enterprise
+  - 0.6b
+library_name: transformers
 ---
 
 # DeepBrainz-R1-0.6B-Exp
 
-**DeepBrainz-R1-0.6B-Exp** is a compact, reasoning-focused language model designed for efficient problem-solving in **mathematics, logic, and code-related tasks**.
+**DeepBrainz-R1-0.6B-Exp** is a compact, high-performance reasoning model engineered by **DeepBrainz AI & Labs**. Designed for efficiency and scalability, it specializes in structured chain-of-thought reasoning, mathematical problem solving, and logical analysis.
 
-Despite its small size, R1-0.6B-Exp emphasizes experimental **structured reasoning**, **stepwise problem decomposition**, and **stable generation behavior**, making it well-suited for research, education, and lightweight deployment scenarios.
+This model is part of the **DeepBrainz-R1 Series**, built to deliver frontier-class reasoning capabilities in cost-effective parameter sizes.
 
 ---
 
-## Model Highlights
+## 🚀 Model Highlights
 
-- Compact **0.6B parameter** model optimized for efficiency
-- Strong focus on **reasoning-oriented tasks**
-- Stable long-form generation for its size class
-- Compatible with standard Hugging Face inference tooling
+- **Parameter Count:** ~0.6B
+- **Context Window:** 32,768 tokens
+- **Specialization:** STEM Reasoning, Logic, Code Analysis
+- **Architecture:** Optimized Dense Transformer (Qwen2.5/3 Compatible)
+- **Deployment:** Ready for vLLM, TGI, and local inference
 
 ---
 
-## Intended Use
+## 🎯 Intended Use Cases
 
-This model is intended for:
-- Research and experimentation in reasoning-focused LLMs
-- Educational use and demonstrations
-- Lightweight inference environments
-- Building blocks for agentic or tool-augmented systems
+- **Agentic Workflows:** Reliability in multi-step planning tasks.
+- **Math & Science:** Solving complex word problems and equations.
+- **Code Generation:** Writing and debugging algorithms.
+- **Structured Data Extraction:** Parsing and reasoning over unstructured text.
 
-It is **not** intended as a general-purpose chat replacement for larger frontier models.
+> **Note:** This is a base reasoning model. For conversational chat, we recommend using a specific instruct template or fine-tuning on your domain data.
 
 ---
 
-## Usage
+## 💻 Usage
 
 ```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
@@ -48,47 +50,34 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 model_id = "DeepBrainz/DeepBrainz-R1-0.6B-Exp"
 
 tokenizer = AutoTokenizer.from_pretrained(model_id)
-model = AutoModelForCausalLM.from_pretrained(model_id)
-
-prompt = "Solve step by step: If x + 3 = 7, what is x?"
-inputs = tokenizer(prompt, return_tensors="pt")
-
-outputs = model.generate(
-    **inputs,
-    max_new_tokens=256,
-    temperature=0.6,
-    top_p=0.95,
-    do_sample=True,
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    torch_dtype="bfloat16",
+    device_map="auto"
 )
 
+prompt = "Analyze the time complexity of the following algorithm:"
+inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+
+outputs = model.generate(**inputs, max_new_tokens=256)
 print(tokenizer.decode(outputs[0], skip_special_tokens=True))
 ```
 
 ---
 
-## Training & Alignment
+## 🛡️ Limitations & Safety
 
-R1-0.6B-Exp was trained using modern post-training techniques emphasizing reasoning quality and generation stability.
-Specific training details are intentionally abstracted in this public-facing release.
+While this model demonstrates strong reasoning capabilities, it may still produce inaccurate information ("hallucinations"). Users should implement appropriate guardrails for production deployments.
 
 ---
 
-## Limitations
+## 📜 License
 
-Performance is constrained by model size
-Not optimized for open-ended conversational chat
-Best for short-to-medium complexity reasoning tasks
+This model is released under the **Apache 2.0** license, allowing for academic and commercial use.
 
 ---
 
-## License
-
-Apache 2.0
-
----
-
-## About DeepBrainz
-
-DeepBrainz builds reasoning-first AI systems focused on efficiency, structure, and real-world problem-solving.
-
-More evaluations and updates will follow in future releases.
+<div align="center">
+  <b>DeepBrainz AI & Labs</b><br>
+  <i>Advancing General Intelligence through Scalable Reasoning</i>
+</div>

config.json

@@ -9,21 +9,10 @@
   "num_attention_heads": 16,
   "num_key_value_heads": 8,
   "head_dim": 128,
-  "max_position_embeddings": 16384,
-  "rms_norm_eps": 1e-06,
-  "rope_theta": 1000000.0,
-  "rope_scaling": null,
-  "attention_bias": false,
-  "attention_dropout": 0.0,
-  "hidden_act": "silu",
-  "initializer_range": 0.02,
-  "tie_word_embeddings": false,
-  "torch_dtype": "bfloat16",
-  "transformers_version": "4.45.0",
-  "use_cache": true,
-  "use_sliding_window": false,
+  "max_position_embeddings": 32768,
   "vocab_size": 151936,
+  "torch_dtype": "bfloat16",
   "bos_token_id": 151643,
   "eos_token_id": 151645,
-  "pad_token_id": 151643
+  "tie_word_embeddings": false
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9920f3b4a9a3280f669b8a51343c16b82fecd48d3c90bb6fda297d50d9602bab
-size 3006565024
+oid sha256:ff1f2c3133974d003a6e654fb47838985161a3b7ca3d85f63a93b668773675e2
+size 1503300296