gpt-oss-20b-lora-finetuned_fp4_step_40

This is a merged model combining GPT-OSS-20B with a fine-tuned LoRA adapter.

Model Details

• Base Model: openai/gpt-oss-20b
• LoRA Checkpoint: checkpoint-40
• Model Type: Causal Language Model
• Model Size: ~20B parameters
• Tensor Type: bfloat16

LoRA Configuration

• Rank (r): 8
• Alpha: 16
• Target Modules: k_proj, v_proj, o_proj, q_proj
• Special MLP Expert Layers: Layers 7, 15, 23

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
    "ChickenMcSwag/gpt-oss-20b-lora-finetuned_fp4_step_40",
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("ChickenMcSwag/gpt-oss-20b-lora-finetuned_fp4_step_40")

# Generate text
prompt = "The future of AI is"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(
    **inputs,
    max_length=100,
    temperature=0.7,
    do_sample=True,
    top_p=0.95
)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

Hardware Requirements

• Minimum VRAM: ~40GB for inference
• Recommended: 2x A100 80GB or equivalent

License

This model follows the original GPT-OSS-20B license. Please refer to the base model's license and usage policy.

Citation

If you use this model, please cite the original GPT-OSS-20B model.