Instructions to use ecloudtech/Erk-32B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ecloudtech/Erk-32B with Transformers:

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

pipe = pipeline("text-generation", model="ecloudtech/Erk-32B")
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("ecloudtech/Erk-32B")
model = AutoModelForCausalLM.from_pretrained("ecloudtech/Erk-32B", 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 ecloudtech/Erk-32B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ecloudtech/Erk-32B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ecloudtech/Erk-32B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ecloudtech/Erk-32B

SGLang

How to use ecloudtech/Erk-32B 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 "ecloudtech/Erk-32B" \
    --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": "ecloudtech/Erk-32B",
		"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 "ecloudtech/Erk-32B" \
        --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": "ecloudtech/Erk-32B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ecloudtech/Erk-32B with Docker Model Runner:
```
docker model run hf.co/ecloudtech/Erk-32B
```

aigencydev commited on Sep 30

Commit

d2679db

verified ·

1 Parent(s): 58a2d4a

egitim/: v2 betikleri ve veri ozetleri

Browse files

Files changed (1) hide show

egitim/birlestir_lineer.py +17 -0

egitim/birlestir_lineer.py ADDED Viewed

	@@ -0,0 +1,17 @@

+# -*- coding: utf-8 -*-
+"""Agirlik uzayinda dogrusal karisim: CIK = a*A + (1-a)*B (Erk-32B'nin 0,75 birlesigiyle ayni yontem).
+Amac: tur-3 modelinin morfoloji kazanimini korurken genel yetenegi (TurkishMMLU) tabana dogru geri cekmek.
+ON-KAYIT: oran DEV + TurkishMMLU ile secilir (A >= 66,18 saglayanlar arasinda DEV farki en yuksek), cekirdek BIR KEZ."""
+import os, sys, torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+A, B, CIK, a = sys.argv[1], sys.argv[2], sys.argv[3], float(sys.argv[4])
+ma = AutoModelForCausalLM.from_pretrained(A, dtype=torch.bfloat16, device_map="cpu")
+mb = AutoModelForCausalLM.from_pretrained(B, dtype=torch.bfloat16, device_map="cpu")
+sb = mb.state_dict(); n = 0
+with torch.no_grad():
+    for k, p in ma.state_dict().items():
+        q = sb[k]
+        assert p.shape == q.shape, k
+        p.copy_((a * p.float() + (1 - a) * q.float()).to(p.dtype)); n += p.numel()
+os.makedirs(CIK, exist_ok=True); ma.save_pretrained(CIK, safe_serialization=True); AutoTokenizer.from_pretrained(A).save_pretrained(CIK)
+print("yazildi %s  a=%.2f  %.2fB parametre" % (CIK, a, n / 1e9), flush=True)