Instructions to use PYTHAI/mindXtrain39 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use PYTHAI/mindXtrain39 with Transformers:

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

pipe = pipeline("text-generation", model="PYTHAI/mindXtrain39")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("PYTHAI/mindXtrain39")
model = AutoModelForCausalLM.from_pretrained("PYTHAI/mindXtrain39", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use PYTHAI/mindXtrain39 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "PYTHAI/mindXtrain39"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PYTHAI/mindXtrain39",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/PYTHAI/mindXtrain39

SGLang

How to use PYTHAI/mindXtrain39 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 "PYTHAI/mindXtrain39" \
    --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": "PYTHAI/mindXtrain39",
		"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 "PYTHAI/mindXtrain39" \
        --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": "PYTHAI/mindXtrain39",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use PYTHAI/mindXtrain39 with Docker Model Runner:
```
docker model run hf.co/PYTHAI/mindXtrain39
```

mindXtrain39

File size: 1,314 Bytes

{
 "routine": "impression.bootcamp",
 "version": 1,
 "direction": "impression → drill: read what the last impression actually taught, and let it write the next drill",
 "input": {
  "probes": "every imprint and coach probe, joined to the corpus row that taught it (by question hash)",
  "attribution": "learned.py rolls probes up per source → probes, recalled, mean accuracy after"
 },
 "rules": [
  "raise the weight of a source whose probes were NOT recalled; lower the weight of one that holds (0.5..3, never drop a row)",
  "if identity < 50% on an accepted imprint, the next drill is persona-dominant — the model is served but not yet mindX",
  "three runs with no positive influence is training_stalled: stop asking for more compute, change the drill",
  "never re-try a rung the ladder evidence rejected (coach_recommend enforces this)"
 ],
 "output": {
  "next_recipe": "suggest_next_params (lr, epochs, LoRA shape, persona share, seq len)",
  "adopt": "POST /hf/coach/recipe {adopt:true} — it travels to the next run",
  "corpus_weights": "data/monitoring/learned_sources.json"
 },
 "evidence_now": {
  "lineage": "37 attempts after generation 39: 30 proof_rejected (41, 46-74), 6 train_failed (40, 42-45, 75), none accepted",
  "reading": "the drill, not the compute, is what has to change"
 }
}