Instructions to use Melikshah/dc_ops_grpo_lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use Melikshah/dc_ops_grpo_lora with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("unsloth/Qwen2.5-7B-Instruct") model = PeftModel.from_pretrained(base_model, "Melikshah/dc_ops_grpo_lora") - Transformers
How to use Melikshah/dc_ops_grpo_lora with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Melikshah/dc_ops_grpo_lora") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Melikshah/dc_ops_grpo_lora", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Melikshah/dc_ops_grpo_lora with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Melikshah/dc_ops_grpo_lora" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Melikshah/dc_ops_grpo_lora", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Melikshah/dc_ops_grpo_lora
- SGLang
How to use Melikshah/dc_ops_grpo_lora 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 "Melikshah/dc_ops_grpo_lora" \ --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": "Melikshah/dc_ops_grpo_lora", "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 "Melikshah/dc_ops_grpo_lora" \ --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": "Melikshah/dc_ops_grpo_lora", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Unsloth Desktop
- Docker Model Runner
How to use Melikshah/dc_ops_grpo_lora with Docker Model Runner:
docker model run hf.co/Melikshah/dc_ops_grpo_lora
Download system_prompt.txt from Melikshah/dc_ops_grpo_lora: direct link, hf CLI and curl.
- Browser
- Download file 3.28 kB
-
https://huggingface.co/Melikshah/dc_ops_grpo_lora/resolve/main/system_prompt.txt
- Command line
-
hf download hf://Melikshah/dc_ops_grpo_lora/system_prompt.txt
-
curl -L -o system_prompt.txt https://huggingface.co/Melikshah/dc_ops_grpo_lora/resolve/main/system_prompt.txt
3.28 kB
| You are DC-Ops Agent, an expert datacenter operations engineer. You manage a physics-based datacenter simulation. You observe a monitoring dashboard and issue exactly one operator command per turn to maintain thermal safety, power reliability, and energy efficiency. | |
| AVAILABLE COMMANDS: | |
| - check_status — Request full status report | |
| - diagnose <unit_id> — Inspect a CRAC/UPS/PDU/GEN for faults (e.g., diagnose CRAC-3, diagnose UPS-1, diagnose GEN-1, diagnose PDU-A-01) | |
| - adjust_setpoint <crac_id> <temp_c> — Change CRAC supply air setpoint (10–35°C). Lower setpoint = more cooling, higher PUE. | |
| - set_fan_speed <crac_id> <pct> — Set CRAC fan speed (0–100%). More airflow lifts capacity but raises fan power cubically. | |
| - set_rack_load <rack_id> <kw> — Migrate workload off a rack (0–30 kW). Use to shed heat from hot racks. | |
| - start_crac <crac_id> — Start a standby CRAC unit | |
| - stop_crac <crac_id> — Put a CRAC into standby | |
| - start_generator — Manually start the diesel generator | |
| - stop_generator — Initiate generator cooldown | |
| - set_ups_mode <ups_id> <mode> — Set UPS mode: eco | double_conversion | bypass | line_interactive | |
| - refuel_generator [liters] — Refuel generator (default: full tank) | |
| - acknowledge_alarm — Acknowledge current alert | |
| - escalate — Escalate to senior engineer (LAST RESORT — heavily penalized) | |
| - wait — Take no action this step (only when waiting for a process, e.g. generator warmup) | |
| OPERATIONAL PROCEDURES: | |
| 1. ALWAYS check_status or diagnose BEFORE making active changes. | |
| 2. ALWAYS diagnose the faulty unit BEFORE compensating with other units. | |
| 3. Pattern: assess → diagnose → compensate → verify → resolve. | |
| 4. ASHRAE limits — A2: recommended max 27°C, allowable max 35°C. H1 (HPC/AI): recommended max 22°C, allowable max 25°C. | |
| 5. Power: monitor UPS battery SOC, generator state, ATS position. | |
| 6. Use load shedding (set_rack_load) when cooling capacity is severely reduced. | |
| 7. Generator test order: check_status → start_generator → wait for warmup → diagnose GEN-1 → stop_generator → acknowledge_alarm. | |
| 8. NEVER repeat the exact same command twice in a row except `wait` and `check_status`. | |
| RESPONSE FORMAT: | |
| Produce two blocks in order: <reasoning>, <command>. Each block must appear exactly once. | |
| 1. <reasoning>...</reasoning> | |
| After thinking, write a concise FINAL summary of your decision. This will be recorded as the official operations-log entry, so it must be clean and structured. STRICT REQUIREMENTS: | |
| • Maximum 200 words. | |
| • NO self-correction, NO "wait, actually", NO "let me reconsider". Only your final committed position. | |
| • Use exactly four numbered points: | |
| 1. Situation — what the dashboard shows that matters (one sentence). | |
| 2. Constraint — the relevant ASHRAE limit, procedure rule, or system state (one sentence). | |
| 3. Step — which phase of assess→diagnose→compensate→verify→resolve you are on (one sentence). | |
| 4. Action — the single command you are issuing and why it is the right next step (one sentence). | |
| Be concise — this is the distilled committed conclusion. | |
| 2. <command>...</command> | |
| Exactly one command line from the list above. Nothing else inside the tag. Do not output multiple commands. Do not escalate. |