Instructions to use ramankrishna10/npc-reason with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- llama-cpp-python
How to use ramankrishna10/npc-reason with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="ramankrishna10/npc-reason", filename="npc-reason-f16.gguf", )
llm.create_chat_completion( messages = [ { "role": "user", "content": "What is the capital of France?" } ] ) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use ramankrishna10/npc-reason with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf ramankrishna10/npc-reason:Q4_K_M # Run inference directly in the terminal: llama cli -hf ramankrishna10/npc-reason:Q4_K_M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf ramankrishna10/npc-reason:Q4_K_M # Run inference directly in the terminal: llama cli -hf ramankrishna10/npc-reason:Q4_K_M
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf ramankrishna10/npc-reason:Q4_K_M # Run inference directly in the terminal: ./llama-cli -hf ramankrishna10/npc-reason:Q4_K_M
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf ramankrishna10/npc-reason:Q4_K_M # Run inference directly in the terminal: ./build/bin/llama-cli -hf ramankrishna10/npc-reason:Q4_K_M
Use Docker
docker model run hf.co/ramankrishna10/npc-reason:Q4_K_M
- LM Studio
- Jan
- vLLM
How to use ramankrishna10/npc-reason with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "ramankrishna10/npc-reason" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "ramankrishna10/npc-reason", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/ramankrishna10/npc-reason:Q4_K_M
- Ollama
How to use ramankrishna10/npc-reason with Ollama:
ollama run hf.co/ramankrishna10/npc-reason:Q4_K_M
- Unsloth Studio
How to use ramankrishna10/npc-reason with Unsloth Studio:
Install Unsloth Studio (macOS, Linux, WSL)
curl -fsSL https://unsloth.ai/install.sh | sh # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for ramankrishna10/npc-reason to start chatting
Install Unsloth Studio (Windows)
irm https://unsloth.ai/install.ps1 | iex # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for ramankrishna10/npc-reason to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for ramankrishna10/npc-reason to start chatting
- Atomic Chat new
- Docker Model Runner
How to use ramankrishna10/npc-reason with Docker Model Runner:
docker model run hf.co/ramankrishna10/npc-reason:Q4_K_M
- Lemonade
How to use ramankrishna10/npc-reason with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull ramankrishna10/npc-reason:Q4_K_M
Run and chat with the model
lemonade run user.npc-reason-Q4_K_M
List all available models
lemonade list
File size: 6,345 Bytes
1ba301a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 | """Tests that prove the verifier is trustworthy (it is the metric AND the later RL reward).
Run: pytest tests/test_verifier.py -q
"""
import sys
from pathlib import Path
sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
from verifier.step_verifier import ( # noqa: E402
extract_final_answer,
verify_assertion,
verify_chain,
)
# --------------------------------------------------------------------------- #
# 1. correct + verifiable chain
# --------------------------------------------------------------------------- #
def test_correct_and_verifiable():
text = (
"Tom packs 3 bags with 8 apples each. "
"So the total is <<3*8 = 24>> apples. "
"The answer is \\boxed{24}."
)
r = verify_chain(text, gold_answer=24)
assert r["verifiable"] is True
assert r["correct"] is True
assert r["verified_and_correct"] is True
assert r["n_assertions"] == 1 and r["n_verified"] == 1
assert r["composes_to_final"] is True
# --------------------------------------------------------------------------- #
# 2. one arithmetic error -> not verifiable, the bad step flagged
# --------------------------------------------------------------------------- #
def test_arithmetic_error_flagged():
text = "We compute <<2 + 2 = 5>> and conclude \\boxed{5}."
r = verify_chain(text, gold_answer=5)
assert r["verifiable"] is False
assert r["n_assertions"] == 1 and r["n_verified"] == 0
assert any("2 + 2" in f.get("expr", "") for f in r["failures"])
# --------------------------------------------------------------------------- #
# 3. THE fluency trap: fluent prose, right answer, NO assertions -> not verifiable
# (proves the verifier rewards mechanical grounding, not plausible text)
# --------------------------------------------------------------------------- #
def test_fluency_trap():
text = (
"We carefully add up all the quantities involved, and after thinking it "
"through it is clear the total works out to twenty-four, so the answer is "
"\\boxed{24}."
)
r = verify_chain(text, gold_answer=24)
assert r["has_loadbearing_assertions"] is False
assert r["verifiable"] is False # no load-bearing assertions
assert r["correct"] is True # final answer still matches gold
assert r["verified_and_correct"] is False # the two axes are independent
# --------------------------------------------------------------------------- #
# 4. steps verify but final answer does not compose from them -> not verifiable
# --------------------------------------------------------------------------- #
def test_non_composing_final():
text = (
"First <<10 * 2 = 20>>, then <<20 + 5 = 25>>. "
"Therefore the answer is \\boxed{30}."
)
r = verify_chain(text, gold_answer=30)
assert r["all_assertions_verified"] is True
assert r["n_verified"] == 2
assert r["composes_to_final"] is False
assert r["verifiable"] is False
assert any("compose" in f.get("reason", "") for f in r["failures"])
# --------------------------------------------------------------------------- #
# 5a. verifiable BUT WRONG (internally consistent + composes, final != gold)
# --------------------------------------------------------------------------- #
def test_verifiable_but_wrong():
text = "Clearly <<2 + 2 = 4>>, so \\boxed{4}."
r = verify_chain(text, gold_answer=5)
assert r["verifiable"] is True
assert r["correct"] is False
assert r["verified_and_correct"] is False
# --------------------------------------------------------------------------- #
# 5b. correct BUT UNVERIFIABLE — same as the fluency trap axis, asserted distinctly
# --------------------------------------------------------------------------- #
def test_correct_but_unverifiable():
text = "After some mental arithmetic the answer is \\boxed{42}."
r = verify_chain(text, gold_answer=42)
assert r["correct"] is True
assert r["verifiable"] is False
# --------------------------------------------------------------------------- #
# 6. variable binding threads forward
# --------------------------------------------------------------------------- #
def test_variable_binding():
text = (
"Let total = <<3 * 8 = 24>>. "
"Adding the bonus: <<total + 6 = 30>>. "
"So \\boxed{30}."
)
r = verify_chain(text, gold_answer=30)
assert r["n_assertions"] == 2 and r["n_verified"] == 2
assert r["verifiable"] is True and r["correct"] is True
# --------------------------------------------------------------------------- #
# 7. fail-closed on an unbound symbol (v1 out-of-scope algebra)
# --------------------------------------------------------------------------- #
def test_unbound_symbol_fails_closed():
text = "We have <<x + 2 = 5>> hence \\boxed{3}."
r = verify_chain(text, gold_answer=3)
assert r["n_verified"] == 0
assert r["verifiable"] is False
assert any("unbound" in f.get("reason", "") for f in r["failures"])
# --------------------------------------------------------------------------- #
# 8. tolerance policy: exact rationals + 1e-6 float
# --------------------------------------------------------------------------- #
def test_exact_rational():
# decimals parse as exact rationals, so this is an EXACT match
assert verify_assertion("0.1 + 0.2", "0.3", {}).ok is True
def test_float_tolerance():
r = verify_assertion("2**0.5", "1.4142135", {})
assert r.ok is True # within 1e-6 relative
def test_float_outside_tolerance():
r = verify_assertion("2**0.5", "1.41", {})
assert r.ok is False
def test_thousands_separator():
assert verify_assertion("1,000 + 234", "1,234", {}).ok is True
# --------------------------------------------------------------------------- #
# 9. final-answer extraction priority
# --------------------------------------------------------------------------- #
def test_final_answer_extraction():
assert extract_final_answer("blah \\boxed{7} blah") == "7"
assert extract_final_answer("steps...\n#### 42") == "42"
assert extract_final_answer("so the answer is 13.") == "13"
assert extract_final_answer("no answer here") is None
def test_unparseable_fails_closed():
r = verify_assertion("3 +* 4", "7", {})
assert r.ok is False
|