Views
No views yet
Note: While the original fine-tuning focused on text-only reasoning tasks, this model inherits multimodal capabilities from the base Qwen3.5-35B-A3B. The vision encoder is preserved and functional for image understanding tasks.
1from vllm import LLM, SamplingParams
2
3llm = LLM(
4 model="codgician/Qwen3.5-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GPTQ-int4",
5 trust_remote_code=True,
6 max_model_len=4096,
7 gpu_memory_utilization=0.9,
8)
9
10sampling_params = SamplingParams(temperature=0.7, max_tokens=2048)
11prompt = "Explain the difference between TCP and UDP protocols."
12outputs = llm.generate([prompt], sampling_params)
13print(outputs[0].outputs[0].text)1from vllm import LLM, SamplingParams
2
3llm = LLM(
4 model="codgician/Qwen3.5-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-GPTQ-int4",
5 trust_remote_code=True,
6 max_model_len=4096,
7 gpu_memory_utilization=0.9,
8)
9
10sampling_params = SamplingParams(temperature=0.7, max_tokens=256)
11messages = [
12 {
13 "role": "user",
14 "content": [
15 {"type": "image_url", "image_url": {"url": "https://example.com/image.jpg"}},
16 {"type": "text", "text": "What is in this image?"}
17 ]
18 }
19]
20outputs = llm.chat(messages, sampling_params)
21print(outputs[0].outputs[0].text)| Precision | VRAM (Approx.) |
|---|---|
| INT4 GPTQ | ~22 GB |