PresenceAI

Calliope-32B-SFT-LoRA-v1.0

README

License: apache-2.0

Model details

Table with columns: Item, Value
Item	Value
Developed by	PresenceAI
Series	Presence Muses
Release	v1.0 (GA)
Model type	LoRA adapter (PEFT)
Training stage	Supervised fine-tuning (SFT)
Base model	`Qwen/Qwen3-32B`
Languages	English, Chinese
License	Apache-2.0
Adapter rank	r = 32, alpha = 32, dropout = 0.0
Target modules	`q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`

Model lineage

text

Qwen/Qwen3-32B  ──SFT──▶  Calliope-32B-SFT-LoRA-v1.0  (this repository)

Trained directly from the base model. No intermediate checkpoint or upstream adapter is required — load this adapter on Qwen/Qwen3-32B as shown below. The base weights are not included in this repository.

Intended use

Multi-turn roleplay and companion-style conversation.
Tool-calling assistants that emit a send_message call with the reply text.
Research on persona consistency and conversational style adaptation.

Out of scope: factual question answering, medical / legal / financial advice, high-stakes decision making, or any use that requires verified knowledge.

Training

Method: supervised fine-tuning (SFT) with LoRA.
Training data: 10,000 internal multi-turn conversations with tool-call annotations. The training data is not released with this adapter.
Maximum sequence length: 22,528 tokens.
Epochs: 3.
Effective batch size: 16 (per-device batch 1 × gradient accumulation 16).
Learning rate: 1e-4.
LoRA: r = 32, alpha = 32, dropout = 0.

Evaluation

Automatic evaluation on an internal held-out split (token-level eval loss):

Table with columns: Step, Eval loss, Note
Step	Eval loss	Note
600	0.6724	Best checkpoint; this repository
1,200	0.6892
1,875	0.8184	Overfitting begins after step ~600

These are automatic loss values only. No task-level benchmark, human preference study, or safety evaluation is included. Evaluate on your own data before deployment.

Usage

PEFT / Transformers

python

from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base_id = "Qwen/Qwen3-32B"adapter_id = "PresenceAI/Calliope-32B-SFT-LoRA-v1.0" tokenizer = AutoTokenizer.from_pretrained(base_id)model = AutoModelForCausalLM.from_pretrained(    base_id, torch_dtype="bfloat16", device_map="auto")model = PeftModel.from_pretrained(model, adapter_id)model.eval()

vLLM

bash

vllm serve Qwen/Qwen3-32B \  --enable-lora --max-lora-rank 32 \  --lora-modules calliope=PresenceAI/Calliope-32B-SFT-LoRA-v1.0 \  --enable-auto-tool-choice --tool-call-parser hermes

The adapter was trained with the Qwen3 chat template and tool-call format:

text

<tool_call>{"name": "send_message", "arguments": {"text": "..."}}</tool_call>

Use the tokenizer and chat_template.jinja included in this repository to keep the prompt format consistent with training.

Limitations

Trained for roleplay/companion conversation; outputs may be stylized, repetitive, or inappropriate outside that domain.
The model may reproduce patterns, names, or phrasing from its training data. The training data is not released; deploy with privacy and safety filters.
Only automatic eval loss is reported. No claim is made about factual accuracy, safety, or human-perceived quality.
The base model's limitations and license terms continue to apply.

License

The adapter is released under the Apache License 2.0. The base model Qwen/Qwen3-32B is also Apache-2.0. See LICENSE and NOTICE in this repository.

Citation

bibtex

@misc{presenceai_calliope_32b_sft_lora_v1,  title  = {PresenceAI Calliope-32B-SFT-LoRA v1.0: roleplay and tool-calling chat},  author = {PresenceAI},  year   = {2026},  note   = {LoRA adapter, Apache-2.0}}

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Model Details

Model Provider

PresenceAI

Model Tree

Base

Qwen/Qwen3-32B

Adapter

this model

Input Modalities

Text

Output Modalities

Text

Supported Functionality

Dedicated Endpoints

Explore FriendliAI today

Get started Talk to an engineer

README

License: apache-2.0

Model details

Table with columns: Item, Value
Item	Value
Developed by	PresenceAI
Series	Presence Muses
Release	v1.0 (GA)
Model type	LoRA adapter (PEFT)
Training stage	Supervised fine-tuning (SFT)
Base model	`Qwen/Qwen3-32B`
Languages	English, Chinese
License	Apache-2.0
Adapter rank	r = 32, alpha = 32, dropout = 0.0
Target modules	`q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`

Model lineage

text

Qwen/Qwen3-32B  ──SFT──▶  Calliope-32B-SFT-LoRA-v1.0  (this repository)

Intended use

Multi-turn roleplay and companion-style conversation.
Tool-calling assistants that emit a send_message call with the reply text.
Research on persona consistency and conversational style adaptation.

Out of scope: factual question answering, medical / legal / financial advice, high-stakes decision making, or any use that requires verified knowledge.

Training

Method: supervised fine-tuning (SFT) with LoRA.
Training data: 10,000 internal multi-turn conversations with tool-call annotations. The training data is not released with this adapter.
Maximum sequence length: 22,528 tokens.
Epochs: 3.
Effective batch size: 16 (per-device batch 1 × gradient accumulation 16).
Learning rate: 1e-4.
LoRA: r = 32, alpha = 32, dropout = 0.

Evaluation

Automatic evaluation on an internal held-out split (token-level eval loss):

Table with columns: Step, Eval loss, Note
Step	Eval loss	Note
600	0.6724	Best checkpoint; this repository
1,200	0.6892
1,875	0.8184	Overfitting begins after step ~600

These are automatic loss values only. No task-level benchmark, human preference study, or safety evaluation is included. Evaluate on your own data before deployment.

Usage

PEFT / Transformers

python

from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base_id = "Qwen/Qwen3-32B"adapter_id = "PresenceAI/Calliope-32B-SFT-LoRA-v1.0" tokenizer = AutoTokenizer.from_pretrained(base_id)model = AutoModelForCausalLM.from_pretrained(    base_id, torch_dtype="bfloat16", device_map="auto")model = PeftModel.from_pretrained(model, adapter_id)model.eval()

from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base_id = "Qwen/Qwen3-32B"adapter_id = "PresenceAI/Calliope-32B-SFT-LoRA-v1.0" tokenizer = AutoTokenizer.from_pretrained(base_id)model = AutoModelForCausalLM.from_pretrained(    base_id, torch_dtype="bfloat16", device_map="auto")model = PeftModel.from_pretrained(model, adapter_id)model.eval()

vLLM

bash

vllm serve Qwen/Qwen3-32B \  --enable-lora --max-lora-rank 32 \  --lora-modules calliope=PresenceAI/Calliope-32B-SFT-LoRA-v1.0 \  --enable-auto-tool-choice --tool-call-parser hermes

vllm serve Qwen/Qwen3-32B \  --enable-lora --max-lora-rank 32 \  --lora-modules calliope=PresenceAI/Calliope-32B-SFT-LoRA-v1.0 \  --enable-auto-tool-choice --tool-call-parser hermes

The adapter was trained with the Qwen3 chat template and tool-call format:

text

<tool_call>{"name": "send_message", "arguments": {"text": "..."}}</tool_call>

Use the tokenizer and chat_template.jinja included in this repository to keep the prompt format consistent with training.

Limitations

Trained for roleplay/companion conversation; outputs may be stylized, repetitive, or inappropriate outside that domain.
The model may reproduce patterns, names, or phrasing from its training data. The training data is not released; deploy with privacy and safety filters.
Only automatic eval loss is reported. No claim is made about factual accuracy, safety, or human-perceived quality.
The base model's limitations and license terms continue to apply.

License

The adapter is released under the Apache License 2.0. The base model Qwen/Qwen3-32B is also Apache-2.0. See LICENSE and NOTICE in this repository.

Citation

bibtex

@misc{presenceai_calliope_32b_sft_lora_v1,  title  = {PresenceAI Calliope-32B-SFT-LoRA v1.0: roleplay and tool-calling chat},  author = {PresenceAI},  year   = {2026},  note   = {LoRA adapter, Apache-2.0}}

@misc{presenceai_calliope_32b_sft_lora_v1,  title  = {PresenceAI Calliope-32B-SFT-LoRA v1.0: roleplay and tool-calling chat},  author = {PresenceAI},  year   = {2026},  note   = {LoRA adapter, Apache-2.0}}