ArunkumarVR commited on
Commit
2e85e10
·
verified ·
1 Parent(s): 5ab9014

Official Release: Optimized & Sanitized

Browse files
Files changed (3) hide show
  1. README.md +37 -48
  2. config.json +3 -14
  3. model.safetensors +2 -2
README.md CHANGED
@@ -4,43 +4,45 @@ language:
4
  - en
5
  pipeline_tag: text-generation
6
  tags:
 
7
  - reasoning
8
- - math
9
- - coding
10
- - distillation
11
- - small-model
 
12
  ---
13
 
14
  # DeepBrainz-R1-0.6B-Exp
15
 
16
- **DeepBrainz-R1-0.6B-Exp** is a compact, reasoning-focused language model designed for efficient problem-solving in **mathematics, logic, and code-related tasks**.
17
 
18
- 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.
19
 
20
  ---
21
 
22
- ## Model Highlights
23
 
24
- - Compact **0.6B parameter** model optimized for efficiency
25
- - Strong focus on **reasoning-oriented tasks**
26
- - Stable long-form generation for its size class
27
- - Compatible with standard Hugging Face inference tooling
 
28
 
29
  ---
30
 
31
- ## Intended Use
32
 
33
- This model is intended for:
34
- - Research and experimentation in reasoning-focused LLMs
35
- - Educational use and demonstrations
36
- - Lightweight inference environments
37
- - Building blocks for agentic or tool-augmented systems
38
 
39
- It is **not** intended as a general-purpose chat replacement for larger frontier models.
40
 
41
  ---
42
 
43
- ## Usage
44
 
45
  ```python
46
  from transformers import AutoModelForCausalLM, AutoTokenizer
@@ -48,47 +50,34 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
48
  model_id = "DeepBrainz/DeepBrainz-R1-0.6B-Exp"
49
 
50
  tokenizer = AutoTokenizer.from_pretrained(model_id)
51
- model = AutoModelForCausalLM.from_pretrained(model_id)
52
-
53
- prompt = "Solve step by step: If x + 3 = 7, what is x?"
54
- inputs = tokenizer(prompt, return_tensors="pt")
55
-
56
- outputs = model.generate(
57
- **inputs,
58
- max_new_tokens=256,
59
- temperature=0.6,
60
- top_p=0.95,
61
- do_sample=True,
62
  )
63
 
 
 
 
 
64
  print(tokenizer.decode(outputs[0], skip_special_tokens=True))
65
  ```
66
 
67
  ---
68
 
69
- ## Training & Alignment
70
 
71
- R1-0.6B-Exp was trained using modern post-training techniques emphasizing reasoning quality and generation stability.
72
- Specific training details are intentionally abstracted in this public-facing release.
73
 
74
  ---
75
 
76
- ## Limitations
77
 
78
- Performance is constrained by model size
79
- Not optimized for open-ended conversational chat
80
- Best for short-to-medium complexity reasoning tasks
81
 
82
  ---
83
 
84
- ## License
85
-
86
- Apache 2.0
87
-
88
- ---
89
-
90
- ## About DeepBrainz
91
-
92
- DeepBrainz builds reasoning-first AI systems focused on efficiency, structure, and real-world problem-solving.
93
-
94
- More evaluations and updates will follow in future releases.
 
4
  - en
5
  pipeline_tag: text-generation
6
  tags:
7
+ - deepbrainz
8
  - reasoning
9
+ - mathematics
10
+ - code
11
+ - enterprise
12
+ - 0.6b
13
+ library_name: transformers
14
  ---
15
 
16
  # DeepBrainz-R1-0.6B-Exp
17
 
18
+ **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.
19
 
20
+ This model is part of the **DeepBrainz-R1 Series**, built to deliver frontier-class reasoning capabilities in cost-effective parameter sizes.
21
 
22
  ---
23
 
24
+ ## 🚀 Model Highlights
25
 
26
+ - **Parameter Count:** ~0.6B
27
+ - **Context Window:** 32,768 tokens
28
+ - **Specialization:** STEM Reasoning, Logic, Code Analysis
29
+ - **Architecture:** Optimized Dense Transformer (Qwen2.5/3 Compatible)
30
+ - **Deployment:** Ready for vLLM, TGI, and local inference
31
 
32
  ---
33
 
34
+ ## 🎯 Intended Use Cases
35
 
36
+ - **Agentic Workflows:** Reliability in multi-step planning tasks.
37
+ - **Math & Science:** Solving complex word problems and equations.
38
+ - **Code Generation:** Writing and debugging algorithms.
39
+ - **Structured Data Extraction:** Parsing and reasoning over unstructured text.
 
40
 
41
+ > **Note:** This is a base reasoning model. For conversational chat, we recommend using a specific instruct template or fine-tuning on your domain data.
42
 
43
  ---
44
 
45
+ ## 💻 Usage
46
 
47
  ```python
48
  from transformers import AutoModelForCausalLM, AutoTokenizer
 
50
  model_id = "DeepBrainz/DeepBrainz-R1-0.6B-Exp"
51
 
52
  tokenizer = AutoTokenizer.from_pretrained(model_id)
53
+ model = AutoModelForCausalLM.from_pretrained(
54
+ model_id,
55
+ torch_dtype="bfloat16",
56
+ device_map="auto"
 
 
 
 
 
 
 
57
  )
58
 
59
+ prompt = "Analyze the time complexity of the following algorithm:"
60
+ inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
61
+
62
+ outputs = model.generate(**inputs, max_new_tokens=256)
63
  print(tokenizer.decode(outputs[0], skip_special_tokens=True))
64
  ```
65
 
66
  ---
67
 
68
+ ## 🛡️ Limitations & Safety
69
 
70
+ While this model demonstrates strong reasoning capabilities, it may still produce inaccurate information ("hallucinations"). Users should implement appropriate guardrails for production deployments.
 
71
 
72
  ---
73
 
74
+ ## 📜 License
75
 
76
+ This model is released under the **Apache 2.0** license, allowing for academic and commercial use.
 
 
77
 
78
  ---
79
 
80
+ <div align="center">
81
+ <b>DeepBrainz AI & Labs</b><br>
82
+ <i>Advancing General Intelligence through Scalable Reasoning</i>
83
+ </div>
 
 
 
 
 
 
 
config.json CHANGED
@@ -9,21 +9,10 @@
9
  "num_attention_heads": 16,
10
  "num_key_value_heads": 8,
11
  "head_dim": 128,
12
- "max_position_embeddings": 16384,
13
- "rms_norm_eps": 1e-06,
14
- "rope_theta": 1000000.0,
15
- "rope_scaling": null,
16
- "attention_bias": false,
17
- "attention_dropout": 0.0,
18
- "hidden_act": "silu",
19
- "initializer_range": 0.02,
20
- "tie_word_embeddings": false,
21
- "torch_dtype": "bfloat16",
22
- "transformers_version": "4.45.0",
23
- "use_cache": true,
24
- "use_sliding_window": false,
25
  "vocab_size": 151936,
 
26
  "bos_token_id": 151643,
27
  "eos_token_id": 151645,
28
- "pad_token_id": 151643
29
  }
 
9
  "num_attention_heads": 16,
10
  "num_key_value_heads": 8,
11
  "head_dim": 128,
12
+ "max_position_embeddings": 32768,
 
 
 
 
 
 
 
 
 
 
 
 
13
  "vocab_size": 151936,
14
+ "torch_dtype": "bfloat16",
15
  "bos_token_id": 151643,
16
  "eos_token_id": 151645,
17
+ "tie_word_embeddings": false
18
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9920f3b4a9a3280f669b8a51343c16b82fecd48d3c90bb6fda297d50d9602bab
3
- size 3006565024
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff1f2c3133974d003a6e654fb47838985161a3b7ca3d85f63a93b668773675e2
3
+ size 1503300296