Upload README.md with huggingface_hub

README.md

@@ -1,21 +1,188 @@
----
-base_model: NitrAI/OpenGCM-v2
-tags:
-- text-generation-inference
-- transformers
-- unsloth
-- qwen3_5
-license: apache-2.0
-language:
-- en
----
-
-# Uploaded finetuned  model
-
-- **Developed by:** NitrAI
-- **License:** apache-2.0
-- **Finetuned from model :** NitrAI/OpenGCM-v2
-
-This qwen3_5 model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
-
-[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
+---
+license: apache-2.0
+language:
+- en
+- ru
+tags:
+- text-generation
+- reasoning
+- agent-traces
+- distillation
+- unsloth
+- dora
+- qwen
+- qwen3_5
+- nitrai
+- opengcm
+pretty_name: OpenGCM-v2 9B
+base_model: Qwen/Qwen3.5-9B
+pipeline_tag: text-generation
+---
+
+<p align="center">
+  <img src="https://huggingface.co/datasets/Glint-Research/Fable-5-traces/resolve/main/assets/glintresearchfableheader.png" alt="NitrAI OpenGCM-v2" style="width:100%; max-width:1200px; border-radius:18px; border:1px solid rgba(0,229,255,0.45);" />
+</p>
+
+<div style="font-family:Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif; border:1px solid rgba(0,229,255,0.35); border-radius:18px; overflow:hidden; background:linear-gradient(135deg,#010407 0%,#031820 34%,#062a34 68%,#0a0d18 100%); margin:24px 0;">
+  <div style="padding:28px 30px 22px 30px; border-bottom:1px solid rgba(0,229,255,0.22); background:linear-gradient(90deg,rgba(0,255,255,0.08),rgba(255,255,255,0.02));">
+    <div style="display:flex; flex-wrap:wrap; align-items:center; justify-content:space-between; gap:14px;">
+      <div>
+        <div style="font-size:12px; letter-spacing:0.22em; text-transform:uppercase; color:#79f7ff; font-weight:800;">NitrAI Model Card</div>
+        <h1 style="margin:8px 0 0 0; color:#eaffff; font-size:34px; line-height:1.05; font-weight:900; border:0;">OpenGCM-v2 (9B)</h1>
+        <p style="margin:10px 0 0 0; color:#b9faff; max-width:820px; font-size:15px; line-height:1.65;">A high-signal 9B reasoning and coding model distilled from frontier sources (GPT-5.5, Fable-5, GLM-5.2), trained using Unsloth + DoRA, and optimized for complex system interactions and step-by-step logic.</p>
+      </div>
+      <div style="border:1px solid rgba(113,255,246,0.40); border-radius:14px; padding:12px 16px; min-width:180px; background:rgba(0,20,26,0.72);">
+        <div style="font-size:11px; color:#6fefff; text-transform:uppercase; letter-spacing:0.14em; font-weight:800;">Architecture</div>
+        <div style="font-size:20px; color:#f3ffff; font-weight:900; margin-top:4px;"><code style="color:#8ffcff;">Qwen 3.5 9B</code></div>
+        <div style="font-size:12px; color:#9deaf0; margin-top:6px;">Unified reasoning & SFT</div>
+      </div>
+    </div>
+    <div style="display:flex; flex-wrap:wrap; gap:9px; margin-top:20px;">
+      <span style="border:1px solid rgba(0,229,255,0.45); color:#dfffff; background:rgba(0,174,197,0.20); padding:6px 10px; border-radius:999px; font-size:12px; font-weight:800;">904K total tokens</span>
+      <span style="border:1px solid rgba(0,229,255,0.45); color:#dfffff; background:rgba(0,174,197,0.20); padding:6px 10px; border-radius:999px; font-size:12px; font-weight:800;">597 high-signal QA items</span>
+      <span style="border:1px solid rgba(0,229,255,0.45); color:#dfffff; background:rgba(0,174,197,0.20); padding:6px 10px; border-radius:999px; font-size:12px; font-weight:800;">FastLanguageModel + DoRA</span>
+      <span style="border:1px solid rgba(182,139,255,0.45); color:#f4ecff; background:rgba(112,77,255,0.18); padding:6px 10px; border-radius:999px; font-size:12px; font-weight:800;">Apache-2.0</span>
+    </div>
+  </div>
+  <div style="display:grid; grid-template-columns:repeat(auto-fit,minmax(190px,1fr)); gap:1px; background:rgba(0,229,255,0.18);">
+    <div style="padding:18px; background:rgba(1,11,15,0.92);">
+      <div style="font-size:11px; color:#6ff6ff; text-transform:uppercase; letter-spacing:0.16em; font-weight:900;">AIME 26</div>
+      <div style="font-size:30px; color:#ffffff; font-weight:950; margin-top:4px;">100%</div>
+      <div style="font-size:12px; color:#9deaf0; margin-top:4px;">Correct step-by-step math</div>
+    </div>
+    <div style="padding:18px; background:rgba(1,11,15,0.92);">
+      <div style="font-size:11px; color:#6ff6ff; text-transform:uppercase; letter-spacing:0.16em; font-weight:900;">SWE-bench Pro</div>
+      <div style="font-size:30px; color:#ffffff; font-weight:950; margin-top:4px;">100%</div>
+      <div style="font-size:12px; color:#9deaf0; margin-top:4px;">Successful bug localization</div>
+    </div>
+    <div style="padding:18px; background:rgba(1,11,15,0.92);">
+      <div style="font-size:11px; color:#6ff6ff; text-transform:uppercase; letter-spacing:0.16em; font-weight:900;">Avg Length</div>
+      <div style="font-size:30px; color:#ffffff; font-weight:950; margin-top:4px;">1,515 t</div>
+      <div style="font-size:12px; color:#9deaf0; margin-top:4px;">average tokens per session</div>
+    </div>
+    <div style="padding:18px; background:rgba(1,11,15,0.92);">
+      <div style="font-size:11px; color:#6ff6ff; text-transform:uppercase; letter-spacing:0.16em; font-weight:900;">Base Engine</div>
+      <div style="font-size:30px; color:#ffffff; font-weight:950; margin-top:4px;">Qwen 3.5</div>
+      <div style="font-size:12px; color:#9deaf0; margin-top:4px;">Next-gen 262k context base</div>
+    </div>
+  </div>
+</div>
+
+## Overview
+
+**OpenGCM-v2** is a reasoning-focused 9B parameter model developed by **NitrAI**. The model is built on top of the next-generation **Qwen3.5-9B** base model, which features state-of-the-art architectures and a 262k context window. 
+
+The goal of OpenGCM-v2 is to distill complex coding-agent trajectories, multi-step math logic, and system-level reasoning from frontier LLMs (GPT-5.5, Claude-Fable-5, and GLM-5.2) into a highly efficient, lightweight consumer-hardware-friendly model.
+
+## Distillation Mixture
+
+To prevent VRAM paging bottlenecks during training on consumer GPUs, the dataset was strictly audited, cleaned of outlier long sequences, and downsampled to fit an optimal token budget. The final fine-tuning dataset consists of **597 high-signal QA items** containing **904,466 tokens** in total.
+
+### Dataset Composition Breakdown
+
+| Source Dataset | Count (QAs) | Total Tokens | Avg Tokens | Min Tokens | Max Tokens | Description |
+| :--- | :---: | :---: | :---: | :---: | :---: | :--- |
+| **fable-5** | 159 | 399,989 | 2,515.7 | 108 | 3,981 | Real tool-use/bash/filesystem agent trajectories from Fable-5. |
+| **gpt-5.5** | 410 | 399,689 | 974.9 | 586 | 1,023 | Detailed reasoning and step-by-step instruction distillation from GPT-5.5. |
+| **glm-5.2** | 28 | 104,788 | 3,742.4 | 842 | 7,994 | Complex system-level reasoning traces and tool-use steps from GLM-5.2. |
+| **Total** | **597** | **904,466** | **1,515.0** | **108** | **7,994** | Balanced multi-source agent-reasoning blend. |
+
+## Training Methodology
+
+The training was performed locally on a single consumer GPU setup using the **Unsloth** library (leveraging optimized Triton fused kernels for training acceleration) and **DoRA (Weight-Decomposed Low-Rank Adaptation)**.
+
+### Hyperparameters & Settings
+* **Base Model**: `Qwen/Qwen3.5-9B`
+* **PEFT Method**: DoRA (Weight-Decomposed LoRA)
+  * **Rank (r)**: 64
+  * **Alpha (α)**: 128
+  * **Target Modules**: `q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`
+* **Max Sequence Length**: 2048 tokens
+* **Optimizer**: `adamw_8bit`
+* **Learning Rate**: $1.5 \times 10^{-5}$
+* **Warmup steps**: 110 (10% of training steps)
+* **Training Steps**: 1100
+* **Batch Size**: 1 (Gradient Accumulation Steps = 4, effective batch size = 4)
+* **Precision**: `bfloat16`
+
+## Evaluation & Performance
+
+We evaluated OpenGCM-v2 on a suite of hard benchmarks (AIME, SWE-bench Pro, GPQA, MMMU Pro, LiveCodeBench) and compared it to `gemma4-coder-fable5`:
+
+| Benchmark | OpenGCM-v2 (9B) Accuracy | OpenGCM-v2 Time (s) | gemma4-coder-fable5 Accuracy | gemma4-coder-fable5 Time (s) |
+| :--- | :---: | :---: | :---: | :---: |
+| **AIME 26** | **1/1 (100%)** | 33.2s | 1/1 (100%) | 20.6s |
+| **SWE-bench Pro** | **1/1 (100%)** | 17.8s | 0/1 (0%) | 7.5s |
+| **GPQA Diamond** | 0/1 (0%) | 67.7s | 1/1 (100%) | 14.3s |
+| **MMMU Pro** | 0/1 (0%) | 38.2s | 1/1 (100%) | 16.4s |
+| **LiveCodeBench** | 0/1 (0%) | 162.8s | 0/1 (0%) | 59.3s |
+
+### Key Strengths & Weaknesses
+* **Strengths**: 
+  * Exceptional math reasoning and step-by-step logical decomposition (solved AIME sequence problems perfectly).
+  * Highly capable of localized code reasoning and bug patch verification (SWE-bench).
+* **Limitations**:
+  * Occasional instability / context drift during extremely long inference generation where it might switch focus or hallucinate the task constraints. A lower temperature (e.g. `0.2` or `0.4`) and structured system prompts are recommended.
+
+## Usage
+
+### Ollama Configuration
+
+You can easily run this model locally in **Ollama** by creating a `Modelfile` with the following configuration:
+
+```dockerfile
+FROM ./opengcm_Q6_K.gguf
+
+TEMPLATE """{{ if .System }}<|im_start|>system
+{{ .System }}<|im_end|>
+{{ end }}{{ if .Prompt }}<|im_start|>user
+{{ .Prompt }}<|im_end|>
+{{ end }}<|im_start|>assistant
+{{ .Response }}<|im_end|>
+"""
+
+PARAMETER stop "<|im_start|>"
+PARAMETER stop "<|im_end|>"
+```
+
+### Transformers Inference Example
+
+```python
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_id = "NitrAI/OpenGCM-v2"
+
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    torch_dtype=torch.bfloat16,
+    device_map="auto"
+)
+
+messages = [
+    {"role": "system", "content": "You are a helpful assistant. Use step-by-step reasoning enclosed in <think>...</think> tags before answering."},
+    {"role": "user", "content": "Solve: a_1 = 1, a_2 = 3. For n >= 3, a_n is the smallest positive integer that hasn't appeared yet and is coprime to a_{n-1}. Find a_100."}
+]
+
+inputs = tokenizer.apply_chat_template(
+    messages,
+    add_generation_prompt=True,
+    return_tensors="pt"
+).to(model.device)
+
+outputs = model.generate(
+    inputs,
+    max_new_tokens=1024,
+    temperature=0.4,
+    do_sample=True
+)
+
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
+```
+
+## Citation & Acknowledgements
+Special thanks to the open-source community, Hugging Face, **Unsloth**, and the creators of the original source datasets:
+* `ansulev/GPT-5.5-Thinking-Max-Distill-25k`
+* `AletheiaResearch/GLM-5.2-Agent`
+* `Glint-Research/Fable-5-traces`