Modelpublic

samuelcardillo/Carnice-MoE-35B-A3B

sourceHugging Faceapache-2.0updated 6mo agoView on Hugging Face

9likes

Model Card

Carnice MoE 35B-A3B — Hermes-Focused Agentic Model

QLoRA fine-tune of Qwen3.5-35B-A3B (MoE, 3B active parameters) optimized for agentic workflows and Hermes Agent runtime. Two-stage training adapted from kai-os/Carnice-9b.

Credits

Training methodology adapted from [kai-os/Carnice-9b](https://huggingface.co/kai-os/Carnice-9b) — same two-stage approach and datasets, applied to the larger MoE architecture. Key inspiration: training on actual Hermes Agent execution traces for native agentic behavior.

For GGUF quantizations (Q4, Q5, Q6, Q8, MXFP4), see samuelcardillo/Carnice-MoE-35B-A3B-GGUF.

Model Details

Property	Value
Base Model	Qwen/Qwen3.5-35B-A3B
Architecture	Mixture of Experts (MoE)
Total Parameters	~35B
Active Parameters	~3B per token

What Makes This Different

Unlike generic reasoning distillation, this model was trained on actual Hermes Agent execution traces — real conversations where an AI agent:

—Executes terminal commands and processes output
—Performs file editing operations
—Chains multi-step tool calls with results feeding back
—Uses browser-assisted workflows
—Makes decisions based on environmental feedback

This teaches the model the exact conversation patterns Hermes expects, rather than just generic reasoning.

Training Details

Two-Stage Approach

Stage A — Reasoning Repair (1 epoch)

—Strengthens base model reasoning before agent-specific training
—Loss: 0.4159

Dataset	Examples
bespokelabs/Bespoke-Stratos-17k	16,710
AI-MO/NuminaMath-CoT	17,000 (capped)

Stage B — Hermes Traces (2 epochs)

—Agent-specific behavioral training on real execution traces
—Loss: 0.3115

Dataset	Examples
kai-os/carnice-glm5-hermes-traces	1,627 (high quality)
open-thoughts/OpenThoughts-Agent-v1-SFT	15,209

Training Configuration

Parameter	Stage A	Stage B
LoRA Rank	64	64
LoRA Alpha	64	64
LoRA Targets	q, k, v, o projections	q, k, v, o projections
Learning Rate	2e-5 (linear)	1e-5 (cosine)
Epochs	1	2
Effective Batch	12	12
Context Length	4096	4096
Precision	4-bit QLoRA + BF16 adapters	Same
GPU	RTX PRO 6000 Blackwell (96GB)	Same
Total Training Time	~44 hours (both stages)

Trainable Parameters

6,881,280 (0.02% of 35B total)

Usage with llama.cpp

bash

llama-server \
  --model Carnice-MoE-35B-A3B-Q8_0.gguf \
  --n-gpu-layers -1 \
  --ctx-size 131072 \
  --host 0.0.0.0 --port 8082

Acknowledgements

—[kai-os](https://huggingface.co/kai-os) — Carnice training methodology and Hermes traces dataset
—[open-thoughts](https://huggingface.co/open-thoughts) — Agent SFT dataset
—[bespokelabs](https://huggingface.co/bespokelabs) — Bespoke-Stratos reasoning dataset
—[Unsloth](https://unsloth.ai) — QLoRA training framework
—[Qwen](https://huggingface.co/Qwen) — Base model