Tanpo Fundraising
A compact founder fundraising specialist (~1.2B) for investor pitches, narrative, ask/use-of-funds, diligence Q&A, updates, and round-structure language (non-legal) — built for local and inexpensive deployment.
Creator: d4rkninja
Collection: Tanpo — Domain Specialists
Original upstream: LiquidAI/LFM2.5-1.2B-Instruct (~1.17B parameters, 32,768-token context, designed for edge/on-device deployment).
Fine-tuning: Unsloth-compatible loading of that checkpoint via hub id unsloth/LFM2.5-1.2B-Instruct (LoRA / PEFT).
This repository hosts the merged Transformers weights (LoRA merged into the base).
Overview
Tanpo is a family of compact domain-specialized business models for local / edge / inexpensive deployment. Different specialists cover different workflows. Tanpo Fundraising is one specialist in that family (not a frontier or general-purpose model).
Related artifacts:
- LoRA adapter: d4rkninja/tanpo-fundraising-LoRA
- GGUF quants: d4rkninja/tanpo-fundraising-GGUF — prefer
Q4_K_Mwhen available
Best For
- Investor pitch openers, narrative arcs, and wedge framing
- Ask + use-of-funds drafts tied to milestones and burn/runway
- Diligence / objection Q&A and investor update drafts
- Traction stories, market/why-now framing, and non-legal round-structure language
Not Designed For
- Legal advice on term sheets, securities, or fiduciary duties
- Investment advice, valuations as fact, or guaranteed raise outcomes
- Unauthorized access, credential attacks, or social-engineering harm
- General coding or non-fundraising chat
Why a Specialist Model?
Fundraising work rewards repeatable structure (narrative, ask, diligence, updates). Specializing a small model for those workflows enables private, low-cost inference without a large general model.
Evaluation
Internal automated domain evaluation (DarkLab harness). Treat as directional, not an industry benchmark.
| Model | Rubric overall |
|---|---|
Base LFM2.5-1.2B-Instruct | 84.2% |
| tanpo-fundraising | 95.2% |
| Delta | +11.1 percentage points |
Delta is calculated from unrounded scores; displayed scores are rounded to one decimal place.
Largest overall gain vs base; trailed on only dilution_round_structure (−22.2 percentage points on that category). Overall still ahead of base; this category is disclosed.
Artifacts: evaluation/ — COMPARE_FUNDRAISING.md, fundraising_tasks.jsonl, score_rubric.md, evaluation/README.md.
Methodology: DarkLab automated domain evaluation (~20 held-out fundraising tasks). Same prompts and generation config for base vs fine-tune. Not an industry benchmark.
Limitations of this eval: Automated rubrics can reward structure over real-world quality; sample size is small; results may not transfer outside the task distribution.
Example Prompts
- User: Rewrite the pitch opener for a B2B SaaS fleet-ops tool moving from seed to Series A bridge. Current story is feature soup — give a sharp narrative arc and talk track.
- User: Build ask + use-of-funds for a Series A raise with explicit burn, runway, and milestone gates to the next raise.
- User: Draft a monthly investor update that kills fluff: KPIs, risks, and concrete intro asks — no customer names.
Inference
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "d4rkninja/tanpo-fundraising"
tokenizer = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo,
torch_dtype="auto",
device_map="auto"
)
messages = [
{"role": "system", "content": "You are Tanpo Fundraising, a practical founder fundraising assistant."},
{"role": "user", "content": "Rewrite the pitch opener for a B2B SaaS fleet-ops tool moving from seed to Series A bridge. Current story is feature soup — give a sharp narrative arc and talk track."},
]
inputs = tokenizer.apply_chat_template(
messages, add_generation_prompt=True, return_tensors="pt"
).to(model.device)
out = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(out[0][inputs.shape[-1]:], skip_special_tokens=True))
Training
Verified from published adapter configs / training artifacts (no unverified hyperparams):
| Field | Value |
|---|---|
| Method | LoRA (PEFT) via Unsloth FastLanguageModel on hub id unsloth/LFM2.5-1.2B-Instruct (Unsloth-compatible of LiquidAI/LFM2.5-1.2B-Instruct) |
LoRA rank (r) | 16 |
| LoRA alpha | 16 |
| LoRA dropout | 0 |
| Bias | none |
| Target modules | Unsloth/PEFT regex targeting attention and MLP projection modules (see adapter_config.json) |
| Task type | CAUSAL_LM |
Merged via PEFT merge_and_unload into full weights in this repo.
Dataset
- d4rkninja/tanpo-fundraising-sft — ~10k format-fixed chat SFT examples (documented on the dataset card).
Limitations
- Specialized: quality drops outside the fundraising workflow distribution.
- ~1.2B scale: limited world knowledge and long-horizon reasoning vs larger models.
- Round-structure / dilution language is draft and non-legal; counsel must review term sheets.
- Eval gains are rubric-based and directional only; one category (
dilution_round_structure) still trails the base.
Responsible Use
Not legal, tax, or investment advice. Humans must review pitches, financial claims, and any term-sheet language. Do not use for fraud, deception, or unauthorized access.
License
license: other / license_name: lfm-1.0
Tanpo merged and GGUF weights are derivatives of LiquidAI/LFM2.5-1.2B-Instruct under the LFM Open License v1.0 (including the commercial Threshold of approximately $10M annual revenue). See the base model card and its LICENSE file. Credit: LiquidAI. Do not treat this stack as Apache-2.0.
Tanpo Family
Tanpo is a family of compact domain-specialized models for focused business workflows.
This specialist is available as:
- Merged:
d4rkninja/tanpo-fundraising - LoRA:
d4rkninja/tanpo-fundraising-LoRA - GGUF:
d4rkninja/tanpo-fundraising-GGUF
Browse all Tanpo specialists: Tanpo — Domain Specialists