0xgr3y/Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Update: Detail opt-use

Browse Source
This commit is contained in:
grey
2025-11-10 08:08:36 +00:00
committed by system
parent fe4f136df4
commit 3cb39a634f
1 changed files with 202 additions and 181 deletions
Download Patch File Download Diff File Expand all files Collapse all files

383
README.md
View File

@@ -17,7 +17,6 @@ tags:
- gguf - gguf
- conversational - conversational
- text-generation-inference - text-generation-inference
- I am tall_tame_panther
pipeline_tag: text-generation pipeline_tag: text-generation
license: apache-2.0 license: apache-2.0
language: language:
@@ -31,16 +30,32 @@ datasets:
- fraction_simplification - fraction_simplification
- basic_arithmetic - basic_arithmetic
inference: true inference: true
widget:
- text: What is 15 * 23?
example_title: Basic Arithmetic
- text: Convert decimal 255 to hexadecimal.
example_title: Base Conversion
- text: Simplify the fraction 24/36.
example_title: Fraction Simplification
model-index: model-index:
- name: Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther - name: Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther
results: [] results:
- task:
type: text-generation
name: Mathematical Reasoning
dataset:
name: Composite Reasoning Dataset
type: custom
metrics:
- type: training_rounds
value: 43610
name: Completed Training Rounds
- type: total_rounds
value: 100000
name: Target Rounds
- type: progress
value: 43.61
name: Training Progress (%)
widget:
- text: "What is 15 * 23?"
example_title: "Basic Arithmetic"
- text: "Convert decimal 255 to hexadecimal."
example_title: "Base Conversion"
- text: "Simplify the fraction 24/36."
example_title: "Fraction Simplification"
--- ---
# Qwen3-0.6B-Gensyn-Swarm (tall_tame_panther) # Qwen3-0.6B-Gensyn-Swarm (tall_tame_panther)
@@ -52,12 +67,12 @@ model-index:
## Model Overview ## Model Overview
This model is a continuously trained Qwen3-0.6B fine-tuned using **Gensyn RL-Swarm** framework with **GRPO (Generalized Reward Policy Optimization)** for enhanced reasoning and mathematical capabilities. This model is a continuously trained Qwen3-0.6B fine-tuned using **Gensyn RL-Swarm** framework with **GRPO (Generalized Reward Policy Optimization)** for enhanced reasoning and mathematical capabilities. **Note: Current training focuses on math/reasoning tasks**.
**Agent ID:** `tall_tame_panther` **Agent ID:** `tall_tame_panther`
**Training Status:** 🔴 LIVE - Model updates automatically every 5-10 minutes **Training Status:** 🟢 LIVE - Model updates automatically every 5-10 minutes
**Current Progress:** Round 43610+ / 1,000,000 **Current Progress:** Round 43,610+ / 100,000 (43,61%)
**Framework Version:** Gensyn RL-Swarm v0.4.2 **Framework Version:** Gensyn RL-Swarm v0.6.4
**Contract:** SwarmCoordinator v0.4.2 **Contract:** SwarmCoordinator v0.4.2
## Key Features ## Key Features
@@ -68,60 +83,24 @@ This model is a continuously trained Qwen3-0.6B fine-tuned using **Gensyn RL-Swa
- **llama.cpp Compatible**: Ready for edge deployment and local inference - **llama.cpp Compatible**: Ready for edge deployment and local inference
- **BF16 Precision**: Trained with bfloat16 for optimal performance - **BF16 Precision**: Trained with bfloat16 for optimal performance
- **TGI Compatible**: Supports Text Generation Inference for production deployment - **TGI Compatible**: Supports Text Generation Inference for production deployment
- **Conversational**: Can be used for interactive reasoning tasks - **Chat Format Support**: Inherits Qwen3 chat template for conversational use
## Training Data ## Training Data
The model is trained on a composite dataset (1,000 samples) with weighted sampling strategy defined in `datasets.yaml`: The model is trained on a composite dataset (1,000 samples) with weighted sampling strategy:
| Dataset | Weight | Samples | Focus Area | | Dataset | Weight | Focus Area |
|---------|--------|---------|------------| |---------|--------|------------|
| Propositional Logic | 7 | 500 | Logical reasoning, truth tables, Boolean operations | | Propositional Logic | 7 | Logical reasoning, truth tables, Boolean operations |
| Calendar Arithmetic | 6 | 500 | Date calculations, leap years, recurring events | | Calendar Arithmetic | 6 | Date calculations, leap years, recurring events |
| Decimal Arithmetic | 5 | 500 | Multi-term decimal operations with precision | | Decimal Arithmetic | 5 | Multi-term decimal operations with precision |
| Base Conversion | 4 | 500 | Number system conversions (base 2-16) | | Base Conversion | 4 | Number system conversions (base 2-16) |
| Fraction Simplification | 4 | 500 | GCD/LCM, fraction reduction | | Fraction Simplification | 4 | GCD/LCM, fraction reduction |
| Basic Arithmetic | 2 | 500 | Foundation operations with parentheses | | Basic Arithmetic | 2 | Foundation operations with parentheses |
**Total Dataset Size:** 1,000 composite samples **Total Dataset Size:** 1,000 composite samples
**Training Samples per Round:** 2 **Training Samples per Round:** 2
**Evaluation Samples:** Real-time via swarm coordination **Evaluation:** Real-time via swarm coordination
### Dataset Configuration Details
```
# From rgym_exp/src/datasets.yaml
Propositional Logic:
- Variables: 2-4
- Statements: 2-4
- Complexity: 1-3
Calendar Arithmetic:
- Year: 2023
- Offset: up to 100 days
- Leap year range: 200 years
- Tasks: count_days, weekday_of_date, is_leap_year, recurring_event_day
Decimal Arithmetic:
- Terms: 2-6
- Decimal places: 1-3
- Precision: 5
Base Conversion:
- Base range: 2-16
- Value range: 0-1000
Fraction Simplification:
- Value range: 1-100
- Factor range: 2-100
- Styles: plain, latex_frac, latex_dfrac
Basic Arithmetic:
- Terms: 2-6
- Digits: 1-4
- Operators: +, -, *, /
- Parentheses: enabled
```
## Quick Start ## Quick Start
@@ -137,13 +116,31 @@ model = AutoModelForCausalLM.from_pretrained(
) )
tokenizer = AutoTokenizer.from_pretrained("0xgr3y/Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther") tokenizer = AutoTokenizer.from_pretrained("0xgr3y/Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther")
# Example: Math reasoning
prompt = "What is 3/4 simplified to lowest terms?" prompt = "What is 3/4 simplified to lowest terms?"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device) inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_length=256, temperature=0.6, top_p=0.95) outputs = model.generate(**inputs, max_length=256, temperature=0.6, top_p=0.95)
print(tokenizer.decode(outputs, skip_special_tokens=True)) print(tokenizer.decode(outputs, skip_special_tokens=True))
``` ```
### Chat Format (Conversational)
```
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("0xgr3y/Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther")
tokenizer = AutoTokenizer.from_pretrained("0xgr3y/Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther")
messages = [
{"role": "system", "content": "You are a helpful math tutor."},
{"role": "user", "content": "Explain how to simplify 24/36 step by step."}
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt")
outputs = model.generate(**inputs, max_length=512)
print(tokenizer.decode(outputs))
```
### Text Generation Inference (TGI) ### Text Generation Inference (TGI)
``` ```
@@ -197,205 +194,239 @@ ollama run qwen3-swarm "What is 15 multiplied by 23?"
All GGUF formats are **llama.cpp compatible** and auto-updated hourly. All GGUF formats are **llama.cpp compatible** and auto-updated hourly.
### GGUF Quantization Strategy
The Q5_K_M format uses mixed precision for optimal quality:
- **Token Embeddings**: Q6_K (high quality vocab representation)
- **Attention Weights**: Q5_K (balanced quality/size)
- **Feed-Forward**: Q5_K/Q6_K (mixed for optimal performance)
- **Layer Norms**: F32 (full precision for stability)
This strategy ensures minimal quality loss while maintaining small file size.
## Chat Format & Conversational Use
This model inherits **Qwen3's chat template** for structured conversations.
### Format Structure
```
<|im_start|>system
{system_message}
<|im_end|>
<|im_start|>user
{user_message}
<|im_end|>
<|im_start|>assistant
{assistant_response}
<|im_end|>
```
### Chat Template Features
- **System Instructions**: Guide model behavior with system messages
- **Multi-turn Dialogue**: Maintains conversation context
- **Tool Calling**: Support function calling (if enabled in training)
- **Reasoning Mode**: `<think>` tags for chain-of-thought (experimental)
**Note**: While the model supports chat format structurally, optimal conversational performance depends on whether training data included formatted dialogues. Current training focuses on **math/reasoning tasks**.
## Training Configuration ## Training Configuration
### Gensyn RL-Swarm Architecture ### Gensyn RL-Swarm Architecture
The model is trained using a decentralized reinforcement learning framework with the following components:
``` ```
# From rgym_exp/config/rg-swarm.yaml
Training Framework: Training Framework:
Method: GRPO (Generalized Reward Policy Optimization) Method: GRPO (Generalized Reward Policy Optimization)
Base Model: Qwen/Qwen3-0.6B Base Model: Qwen/Qwen3-0.6B
Training Regime: bfloat16 mixed precision Training Regime: bfloat16 mixed precision
Max Rounds: 1,000,000 Max Rounds: 100,000
Max Stage: 1
Update Frequency: Every 5-10 minutes Update Frequency: Every 5-10 minutes
Generations per Round: 2 Generations per Round: 2
Transplant Trees: 1
Seed: 42 Seed: 42
Blockchain Integration: Blockchain Integration:
Network: Gensyn Testnet Network: Gensyn Testnet
Chain ID: 685685 Chain ID: 685685
RPC: https://gensyn-testnet.g.alchemy.com/public
Contract: SwarmCoordinator v0.4.2 Contract: SwarmCoordinator v0.4.2
Modal Proxy: http://localhost:3000/api/
Swarm Communication: Swarm Communication:
Framework: Hivemind P2P Backend Framework: Hivemind P2P Backend
Initial Peers: 3 bootnodes Initial Peers: 3 bootnodes
Bootnodes: Beam Size: 30
- /ip4/38.101.215.12/tcp/30011/p2p/QmQ2gEXoPJg6iMBSUFWGzAabS2VhnzuS782Y637hGjfsRJ
- /ip4/38.101.215.13/tcp/30012/p2p/QmWhiaLrx3HRZfgXc2i7KW5nMUNK7P9tRc71yFJdGEZKkC
- /ip4/38.101.215.14/tcp/30013/p2p/QmQa1SCfYTxx7RvU7qJJRo79Zm1RAwPpkeLueDVJuBBmFp
Startup Timeout: 120s
Beam Size: 25
Reward System: Reward System:
Manager: DefaultRewardManager Manager: DefaultRewardManager
Function Store: RoundRewardFnStore Reward Function: RGRewards (Reasoning Gym)
Reward Function: RGRewards (Reasoning Gym Rewards) Judge API: https://swarm-judge.internal-apps-central1.clusters.gensyn.ai
Judge: Swarm Judge API (https://swarm-judge.internal-apps-central1.clusters.gensyn.ai)
``` ```
### Training Hyperparameters ### Model Hyperparameters
``` ```
Model Architecture: Architecture:
Hidden Size: 1024 Hidden Size: 1024
Intermediate Size: 3072 Intermediate Size: 3072
Num Hidden Layers: 28 Layers: 28
Num Attention Heads: 16 Attention Heads: 16
Num Key-Value Heads: 8 KV Heads: 8
Head Dimension: 128 Head Dimension: 128
Max Position Embeddings: 40,960 Context Length: 40,960 tokens
RMS Norm Epsilon: 1e-06 Vocabulary: 151,936 tokens
Rope Theta: 1,000,000
Vocabulary Size: 151,936
GRPO Trainer Config: GRPO Config:
Epsilon: 0.2 Epsilon: 0.2
Epsilon High: 0.28 Epsilon High: 0.28
Generations: 2
Gradient Checkpointing: Enabled Gradient Checkpointing: Enabled
Learning Rate: Adaptive
Generation Config: Generation:
Temperature: 0.6 Temperature: 0.6
Top-K: 20 Top-K: 20
Top-P: 0.95 Top-P: 0.95
BOS Token: 151643
EOS Token: 151645
Pad Token: 151643
``` ```
## Model Capabilities ## Model Capabilities
This model excels at: This model excels at:
1. **Logical Reasoning**: Propositional logic, truth evaluation, Boolean algebra, logical equivalences 1. **Logical Reasoning**: Propositional logic, truth evaluation, Boolean algebra
2. **Mathematical Operations**: Multi-precision arithmetic, decimal calculations, fraction manipulation 2. **Mathematical Operations**: Multi-precision arithmetic, decimal calculations, fractions
3. **Number Systems**: Base conversion between binary, octal, decimal, hexadecimal 3. **Number Systems**: Base conversion (binary, octal, decimal, hexadecimal)
4. **Date/Time Calculations**: Calendar arithmetic, leap year detection, day-of-week calculations 4. **Date/Time Calculations**: Calendar arithmetic, leap years, day-of-week
5. **Step-by-step Problem Solving**: Chain-of-thought reasoning for complex multi-step tasks 5. **Step-by-step Problem Solving**: Chain-of-thought reasoning
6. **Conversational Math Tutoring**: Interactive problem-solving guidance 6. **Conversational Tutoring**: Interactive problem-solving (via chat format)
## Limitations ## Limitations
- **Specialized Domain**: Optimized for reasoning/math tasks; may underperform on creative writing or general chat - **Specialized Domain**: Optimized for reasoning/math; may underperform on creative writing
- **Training in Progress**: Model weights update every 5-10 minutes; performance may vary between checkpoints - **Training in Progress**: Weights update every 5-10 minutes; performance varies
- **Scale**: 0.6B parameters - suitable for edge devices but not state-of-the-art for complex reasoning - **Scale**: 0.6B parameters - suitable for edge but not SOTA for complex reasoning
- **Experimental**: Trained via decentralized RL swarm; behavior may be less predictable than supervised models - **Experimental**: Decentralized RL training; behavior less predictable than supervised models
- **Context Length**: 40K tokens supported but best performance within 4K tokens - **Context**: Best performance within 4K tokens (full 40K supported)
## Update Schedule ## Update Schedule
| Format | Update Frequency | Trigger | | Format | Frequency | Trigger |
|--------|------------------|---------| |--------|-----------|---------|
| Safetensors (BF16) | Every 5-10 minutes | Automatic via RL-Swarm training | | Safetensors (BF16) | Every 5-10 min | Automatic via RL-Swarm |
| GGUF variants (all) | Every 1 hour | Automatic conversion from latest checkpoint | | GGUF (all formats) | Every 1 hour | Auto-conversion pipeline |
**Auto-Conversion Pipeline:** **Auto-Conversion Pipeline:**
- Monitors repo for new training commits 1. Monitors repo for new training commits
- Downloads latest `model.safetensors` 2. Downloads latest `model.safetensors`
- Converts to F16 GGUF base 3. Converts to F16 GGUF base
- Quantizes to Q3_K_M, Q4_K_M, Q5_K_M 4. Quantizes to Q3_K_M, Q4_K_M, Q5_K_M
- Uploads all formats to repo 5. Uploads all formats
Check commit history for exact timestamps of each update. Check commit history for exact timestamps.
## Gensyn RL-Swarm Technical Details ## Gensyn RL-Swarm Technical Details
This model is trained using [Gensyn RL-Swarm](https://gensyn.ai), a decentralized reinforcement learning framework:
### Architecture Components ### Architecture Components
1. **Game Manager** (`rgym_exp/src/manager.py`): Orchestrates training rounds and swarm coordination 1. **Game Manager**: Orchestrates training rounds and swarm coordination
2. **Trainer** (`rgym_exp/src/trainer.py`): GRPO implementation for policy optimization 2. **Trainer**: GRPO implementation for policy optimization
3. **Data Manager** (`rgym_exp/src/data.py`): Handles dataset loading and sampling 3. **Data Manager**: Dataset loading and weighted sampling
4. **Reward Manager** (`rgym_exp/src/rewards.py`): Computes rewards using judge API 4. **Reward Manager**: Computes rewards via judge API
5. **Coordinator** (`rgym_exp/src/coordinator.py`): Blockchain integration for swarm state 5. **Coordinator**: Blockchain integration for swarm state
6. **Communication Backend**: Hivemind DHT for peer-to-peer model sharing 6. **P2P Backend**: Hivemind DHT for model sharing
### Training Process ### Training Process
``` ```
1. Agent joins swarm via P2P network 1. Agent joins swarm via P2P network
2. Coordinator assigns training round via smart contract 2. Coordinator assigns round via smart contract
3. Agent samples data from weighted datasets 3. Agent samples data from weighted datasets
4. Model generates responses (2 generations) 4. Model generates 2 responses
5. Judge API evaluates quality and assigns rewards 5. Judge API evaluates and assigns rewards
6. GRPO updates policy based on rewards 6. GRPO updates policy based on rewards
7. Updated model shared via DHT to swarm 7. Updated model shared via DHT
8. Best model checkpoint saved to HuggingFace 8. Best checkpoint saved to HuggingFace
9. Repeat for next round 9. Repeat
``` ```
### Decentralization Benefits ### Decentralization Benefits
- **Fault Tolerance**: Multiple agents contribute; single node failure doesn't stop training - **Fault Tolerance**: Multiple agents; no single point of failure
- **Diverse Exploration**: Different agents explore different strategies - **Diverse Exploration**: Different agents explore different strategies
- **Collective Intelligence**: Agents learn from each other's experiences - **Collective Intelligence**: Agents learn from each other
- **Transparent Verification**: All training rounds verified on-chain - **Transparent**: All rounds verified on-chain
**Swarm Agent:** `tall_tame_panther` **Swarm Agent:** `tall_tame_panther`
**Contract:** SwarmCoordinator v0.4.2 **Contract:** SwarmCoordinator v0.4.2
**Testnet Explorer:** https://gensyn-testnet.explorer.com
## Technical Specifications ## Technical Specifications
### Software Stack ### Software Stack
- **Training Framework**: Gensyn RL-Swarm v0.4.2 - **Framework**: Gensyn RL-Swarm v0.6.4
- **Base Library**: transformers v4.51.3 - **Library**: transformers v4.51+
- **Communication**: hivemind (P2P backend) - **P2P**: hivemind
- **Blockchain**: Web3.py (Gensyn testnet) - **Blockchain**: Gensyn testnet
- **Configuration**: Hydra + OmegaConf - **Config**: Hydra + OmegaConf
- **Logging**: WandB integration - **Logging**: WandB integration
### Hardware Requirements ### Hardware Requirements
**Training Node:** **Training GPU:**
- GPU: NVIDIA A100 40GB or equivalent (for BF16 training) - GPU: NVIDIA 4090 24GB+ (BF16 training)
- RAM: 32GB+ system memory - RAM: 16GB+
- Cores: 10+
- Storage: 50GB SSD - Storage: 50GB SSD
- Network: High bandwidth for P2P swarm communication - Network: High bandwidth for P2P
**Training CPU Optimize:**
- CPU: INTEL or AMD
- Cores: 10+
- RAM: 16GB+
- Storage: 50GB SSD
- Network: High bandwidth for P2P
**Inference:** **Inference:**
- Safetensors: 8GB+ VRAM (GPU), 16GB+ RAM (CPU) - Safetensors: 8GB VRAM (GPU) / 16GB RAM (CPU)
- GGUF Q4_K_M: 4GB RAM (CPU), 2GB VRAM (GPU) - GGUF Q4_K_M: 2GB VRAM (GPU) / 4GB RAM (CPU)
- GGUF Q3_K_M: 3GB RAM (CPU-only compatible) - GGUF Q3_K_M: 3GB RAM (CPU-only)
## Evaluation
### Training Progress Metrics
| Metric | Value | Target |
|--------|-------|--------|
| Completed Rounds | 43,610+ | 100,000 |
| Training Progress | 43.61% | 100% |
| Update Frequency | 5-10 min | Continuous |
**Note**: Formal evaluation benchmarks (GSM8K, MATH, etc.) will be added as training progresses. Current metrics track training rounds completed in the decentralized swarm.
## Reproducibility ## Reproducibility
To reproduce training results: To reproduce training:
1. Clone Gensyn RL-Swarm repository 1. Clone Gensyn RL-Swarm repository
2. Install dependencies: `pip install -r requirements.txt` 2. Install: `pip install -r requirements.txt`
3. Configure `rgym_exp/config/rg-swarm.yaml` with your settings 3. Configure `rgym_exp/config/rg-swarm.yaml`
4. Set environment variables: 4. Configure `rgym_exp/src/datasets.yaml`
``` 5. Set environment variables:
export HUGGINGFACE_ACCESS_TOKEN=<your-token> ```
export MODEL_NAME=Qwen/Qwen3-0.6B export HUGGINGFACE_ACCESS_TOKEN=<token>
export ORG_ID=<your-org-id> export MODEL_NAME=Qwen/Qwen3-0.6B
export SWARM_CONTRACT=<contract-address> export ORG_ID=<org-id>
``` export SWARM_CONTRACT=<contract-address>
5. Run: `bash run_rl_swarm.sh` ```
6. Run: `bash run_rl_swarm.sh`
**Note:** Exact reproduction requires same seed (42), dataset configuration, and swarm coordination state. **Note**: Exact reproduction requires same seed (42), dataset config, and swarm state.
## Citation ## Citation
``` ```
@misc{qwen3-gensyn-swarm-2025, @misc{qwen3-gensyn-swarm-2025,
author = {0xgr3y}, author = {0xgrey},
title = {Qwen3-0.6B-Gensyn-Swarm: Continuous RL Training on Distributed Swarm}, title = {Qwen3-0.6B-Gensyn-Swarm: Continuous RL Training on Distributed Swarm},
year = {2025}, year = {2025},
publisher = {HuggingFace}, publisher = {HuggingFace},
journal = {HuggingFace Model Hub},
howpublished = {\url{https://huggingface.co/0xgr3y/Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther}}, howpublished = {\url{https://huggingface.co/0xgr3y/Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther}},
note = {Agent ID: tall\_tame\_panther} note = {Agent ID: tall\_tame\_panther}
} }
@@ -404,42 +435,33 @@ To reproduce training results:
title = {Gensyn RL-Swarm: Decentralized Reinforcement Learning Framework}, title = {Gensyn RL-Swarm: Decentralized Reinforcement Learning Framework},
author = {Gensyn AI}, author = {Gensyn AI},
year = {2025}, year = {2025},
url = {https://gensyn.ai}, url = {https://gensyn.ai}
note = {SwarmCoordinator v0.4.2}
}
@article{lacoste2019quantifying,
title={Quantifying the Carbon Emissions of Machine Learning},
author={Lacoste, Alexandre and others},
journal={arXiv preprint arXiv:1910.09700},
year={2019}
} }
``` ```
## References ## References
- **arXiv:1910.09700** - ML Carbon Emissions methodology - **Gensyn Documentation**: https://docs.gensyn.ai/
- **Gensyn Documentation**: https://docs.gensyn.ai - **Gensyn GitHub**: https://github.com/gensyn-ai
- **RL-Swarm Contracts**: https://github.com/gensyn-ai/rl-swarm-contracts
- **Qwen3 Model Card**: https://huggingface.co/Qwen/Qwen3-0.6B - **Qwen3 Model Card**: https://huggingface.co/Qwen/Qwen3-0.6B
- **Technical Report**: See `technical_report.pdf` in training repository - **arXiv:1910.09700**: ML Carbon Emissions methodology
## License ## License
Apache 2.0 - See [LICENSE](LICENSE) for details Apache 2.0 - See [LICENSE](LICENSE)
## Contact & Support ## Contact
- **Developer**: 0xgr3y - **Developer**: 0xgrey
- **Agent ID**: tall_tame_panther - **Agent ID**: tall_tame_panther
- **Issues**: Open an issue on this repo
- **Community**: [Gensyn Discord](https://discord.gg/gensyn) - **Community**: [Gensyn Discord](https://discord.gg/gensyn)
--- ---
**⚠️ Important Note**: This is a continuously trained model. For reproducibility, always specify the exact commit hash: **⚠️ Important**: This is a continuously trained model. For reproducibility, specify commit hash:
``` ```
# Download specific checkpoint
git clone https://huggingface.co/0xgr3y/Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther git clone https://huggingface.co/0xgr3y/Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther
cd Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther cd Qwen3-0.6B-Gensyn-Swarm-tall_tame_panther
git checkout <commit-hash> git checkout <commit-hash>
@@ -453,5 +475,4 @@ git checkout <commit-hash>
[![Gensyn](https://img.shields.io/badge/Powered%20by-Gensyn%20AI-orange?style=for-the-badge)](https://gensyn.ai) [![Gensyn](https://img.shields.io/badge/Powered%20by-Gensyn%20AI-orange?style=for-the-badge)](https://gensyn.ai)
</div> </div>
```