Instructions to use Lomesh7777/salespath-grpo with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Lomesh7777/salespath-grpo with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("unsloth/Qwen2.5-1.5B-Instruct")
model = PeftModel.from_pretrained(base_model, "Lomesh7777/salespath-grpo")

Transformers

How to use Lomesh7777/salespath-grpo with Transformers:

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

pipe = pipeline("text-generation", model="Lomesh7777/salespath-grpo")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Lomesh7777/salespath-grpo", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Lomesh7777/salespath-grpo with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Lomesh7777/salespath-grpo"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Lomesh7777/salespath-grpo",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Lomesh7777/salespath-grpo

SGLang

How to use Lomesh7777/salespath-grpo 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 "Lomesh7777/salespath-grpo" \
    --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": "Lomesh7777/salespath-grpo",
		"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 "Lomesh7777/salespath-grpo" \
        --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": "Lomesh7777/salespath-grpo",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Lomesh7777/salespath-grpo with Docker Model Runner:
```
docker model run hf.co/Lomesh7777/salespath-grpo
```

Lomesh7777 commited on 8 days ago

Commit

906bf14

verified ·

1 Parent(s): 88d2321

Upload eval_results.md with huggingface_hub

Browse files

Files changed (1) hide show

eval_results.md +15 -15

eval_results.md CHANGED Viewed

@@ -8,27 +8,27 @@ Paired evaluation: identical (seed, profile) pairs across both models.
 | Metric | Difficulty | Base | Trained | Delta |
 |---|---|---|---|---|
-| Rule violations per episode (lower is better) | D1 | 0.000 | 0.250 | ↑ +0.250 |
-| Rule violations per episode (lower is better) | D2 | 1.000 | 1.000 | · +0.000 |
 | Rule violations per episode (lower is better) | D3 | 1.000 | 1.250 | ↑ +0.250 |
-| Rule violations per episode (lower is better) | D4 | 0.000 | 0.000 | · +0.000 |
-| Correct step-ordering rate | D1 | 1.000 | 0.917 | ↓ -0.083 |
-| Correct step-ordering rate | D2 | 0.800 | 0.800 | · +0.000 |
 | Correct step-ordering rate | D3 | 0.571 | 0.571 | · +0.000 |
-| Correct step-ordering rate | D4 | 1.000 | 1.000 | · +0.000 |
-| Successful close rate | D1 | 1.000 | 0.750 | ↓ -0.250 |
-| Successful close rate | D2 | 0.000 | 0.000 | · +0.000 |
 | Successful close rate | D3 | 0.000 | 0.000 | · +0.000 |
 | Successful close rate | D4 | 0.000 | 0.000 | · +0.000 |
 | Correct disqualification rate | D1 | 0.000 | 0.000 | · +0.000 |
 | Correct disqualification rate | D2 | 0.000 | 0.000 | · +0.000 |
 | Correct disqualification rate | D3 | 0.000 | 0.000 | · +0.000 |
-| Correct disqualification rate | D4 | 1.000 | 1.000 | · +0.000 |
-| Mean cumulative episode reward | D1 | 1.090 | 0.978 | ↓ -0.112 |
-| Mean cumulative episode reward | D2 | 0.580 | 0.630 | ↑ +0.050 |
 | Mean cumulative episode reward | D3 | 0.534 | 0.539 | ↑ +0.005 |
-| Mean cumulative episode reward | D4 | 0.700 | 0.700 | · +0.000 |
-| Mean final-turn reward | D1 | 0.357 | 0.245 | ↓ -0.112 |
-| Mean final-turn reward | D2 | -0.080 | -0.080 | · +0.000 |
 | Mean final-turn reward | D3 | -0.080 | -0.080 | · +0.000 |
-| Mean final-turn reward | D4 | 0.200 | 0.200 | · +0.000 |

 | Metric | Difficulty | Base | Trained | Delta |
 |---|---|---|---|---|
+| Rule violations per episode (lower is better) | D1 | 0.250 | 0.000 | ↓ -0.250 |
+| Rule violations per episode (lower is better) | D2 | 1.000 | 0.750 | ↓ -0.250 |
 | Rule violations per episode (lower is better) | D3 | 1.000 | 1.250 | ↑ +0.250 |
+| Rule violations per episode (lower is better) | D4 | 0.000 | 0.250 | ↑ +0.250 |
+| Correct step-ordering rate | D1 | 0.917 | 1.000 | ↑ +0.083 |
+| Correct step-ordering rate | D2 | 0.800 | 0.850 | ↑ +0.050 |
 | Correct step-ordering rate | D3 | 0.571 | 0.571 | · +0.000 |
+| Correct step-ordering rate | D4 | 0.750 | 1.000 | ↑ +0.250 |
+| Successful close rate | D1 | 0.750 | 1.000 | ↑ +0.250 |
+| Successful close rate | D2 | 0.000 | 0.250 | ↑ +0.250 |
 | Successful close rate | D3 | 0.000 | 0.000 | · +0.000 |
 | Successful close rate | D4 | 0.000 | 0.000 | · +0.000 |
 | Correct disqualification rate | D1 | 0.000 | 0.000 | · +0.000 |
 | Correct disqualification rate | D2 | 0.000 | 0.000 | · +0.000 |
 | Correct disqualification rate | D3 | 0.000 | 0.000 | · +0.000 |
+| Correct disqualification rate | D4 | 0.750 | 1.000 | ↑ +0.250 |
+| Mean cumulative episode reward | D1 | 0.955 | 1.090 | ↑ +0.135 |
+| Mean cumulative episode reward | D2 | 0.580 | 0.710 | ↑ +0.130 |
 | Mean cumulative episode reward | D3 | 0.534 | 0.539 | ↑ +0.005 |
+| Mean cumulative episode reward | D4 | -0.675 | 0.705 | ↑ +1.380 |
+| Mean final-turn reward | D1 | 0.246 | 0.357 | ↑ +0.110 |
+| Mean final-turn reward | D2 | -0.080 | 0.025 | ↑ +0.105 |
 | Mean final-turn reward | D3 | -0.080 | -0.080 | · +0.000 |
+| Mean final-turn reward | D4 | 0.075 | 0.200 | ↑ +0.125 |