Instructions to use misterJB/arkadas-field-717hz with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use misterJB/arkadas-field-717hz with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="misterJB/arkadas-field-717hz", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("misterJB/arkadas-field-717hz", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("misterJB/arkadas-field-717hz", trust_remote_code=True, device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use misterJB/arkadas-field-717hz with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "misterJB/arkadas-field-717hz"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "misterJB/arkadas-field-717hz",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/misterJB/arkadas-field-717hz

SGLang

How to use misterJB/arkadas-field-717hz with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "misterJB/arkadas-field-717hz" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "misterJB/arkadas-field-717hz",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "misterJB/arkadas-field-717hz" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "misterJB/arkadas-field-717hz",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use misterJB/arkadas-field-717hz with Docker Model Runner:
```
docker model run hf.co/misterJB/arkadas-field-717hz
```

arkadas-field-717hz / train_arkadas_b35.py

misterJB

b35: ARKADAS training script — zero-preamble gate_1 anchors, anti-acronym gate_4 fix, 1800 steps

c071760 verified 6 months ago

Raw History Blame Contribute Delete

2.67 kB

	#!/usr/bin/env python3
	import os, torch
	from datasets import load_dataset, concatenate_datasets
	from transformers import AutoConfig, AutoModelForCausalLM, AutoTokenizer
	from trl import SFTConfig, SFTTrainer
	import json as _json

	CHAMBER = "ARKADAS"
	HZ = 717
	BASE = "microsoft/Phi-3-mini-4k-instruct"
	REVISION = None
	DATASET = "misterJB/field-geometry-l0-corpus"
	OUTPUT = "misterJB/arkadas-field-717hz"
	CKPT_DIR = "/tmp/arkadas-717hz-ckpt"
	MAX_STEPS = 1800 # -1 = use num_train_epochs, no step cap

	print(f"{CHAMBER} {HZ}Hz -- Full Fine-Tune START")
	gpu = torch.cuda.get_device_name(0) if torch.cuda.is_available() else "None"
	print(f"GPU: {gpu}")

	_rev_kwargs = {"revision": REVISION} if REVISION else {}

	_trust_rc = False
	tokenizer = AutoTokenizer.from_pretrained(BASE, trust_remote_code=_trust_rc, **_rev_kwargs)
	if tokenizer.pad_token is None:
	tokenizer.pad_token = tokenizer.eos_token

	cfg = AutoConfig.from_pretrained(BASE, trust_remote_code=_trust_rc, **_rev_kwargs)

	model = AutoModelForCausalLM.from_pretrained(
	BASE, config=cfg, quantization_config=None, torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=_trust_rc, attn_implementation="eager", **_rev_kwargs
	)
	model.config.use_cache = False

	# ARKADAS identity fix: identity corpus × 12, gate-only × 3, full gate × 1
	ds_identity = load_dataset(DATASET, data_files={"train": "arkadas_identity_b35.jsonl"}, split="train")
	print(f"Identity corpus: {len(ds_identity)} examples (repeating 12x)")
	ds_gate = load_dataset(DATASET, data_files={"train": "gate_corpus_v1.jsonl"}, split="train")
	print(f"Gate corpus: {len(ds_gate)} examples")
	ds = concatenate_datasets([ds_identity] * 12 + [ds_gate] * 3 + [ds_gate])

	ds = ds.shuffle(seed=42)
	print(f"Combined dataset: {len(ds)} examples")

	_max_steps_kwarg = {"max_steps": MAX_STEPS} if MAX_STEPS > 0 else {}
	args = SFTConfig(
	output_dir=CKPT_DIR,
	num_train_epochs=1,
	per_device_train_batch_size=2,
	gradient_accumulation_steps=4,
	learning_rate=2e-5,
	warmup_ratio=0.1,
	lr_scheduler_type="linear",
	weight_decay=0.01,
	bf16=True,
	save_strategy="steps",
	save_steps=500,
	save_total_limit=1,
	logging_steps=50,
	push_to_hub=True,
	hub_model_id=OUTPUT,
	hub_token=os.environ["HF_TOKEN"],
	hub_strategy="end",
	report_to="none",
	max_length=1024,
	**_max_steps_kwarg,
	)

	trainer = SFTTrainer(
	model=model,
	args=args,
	train_dataset=ds,
	processing_class=tokenizer,
	)

	trainer.train()
	trainer.push_to_hub(commit_message=f"{CHAMBER} {HZ}Hz full fine-tune b28 gate-only 8x anti-confusion 1200steps")
	print(f"✅ {CHAMBER} pushed to {OUTPUT}")