FP8 Quantized support (#21)

- FP8 Quantized support (e1e295a9b18384283964f7b465315d597d97bc53)

README.md

@@ -22,6 +22,7 @@ pipeline_tag: image-text-to-text
 ## 📢 News & Updates
 
 - 🚀 **Online Demo**: Explore Step3-VL-10B on [Hugging Face Spaces](https://huggingface.co/spaces/stepfun-ai/Step3-VL-10B) !
+- 📢 **[Notice] FP8 Quantization Support :** FP8 quantized weights are now available. ([Download link](https://huggingface.co/stepfun-ai/Step3-VL-10B-FP8))
 - 📢 **[Notice] vLLM Support:** vLLM integration is now officially supported! (PR [#32329](https://github.com/vllm-project/vllm/pull/32329))
 - ✅ **[Fixed] HF Inference:** Resolved the `eos_token_id` misconfiguration in `config.json` that caused infinite generation loops. (PR [#abdf3](https://huggingface.co/stepfun-ai/Step3-VL-10B/commit/abdf3618e914a9e3de0ad74efacc8b7a10f06c10))
 - ✅ **[Fixing] Metric Correction:** We sincerely apologize for inaccuracies in the Qwen3VL-8B benchmarks (e.g., AIME, HMMT, LCB). The errors were caused by an incorrect max_tokens setting (mistakenly set to 32k) during our large-scale evaluation process. We are re-running the tests and will provide corrected numbers in the next version of technical report.
@@ -46,6 +47,7 @@ The success of STEP3-VL-10B is driven by two key strategic designs:
 | :-------------------- | :--- | :----------------------------------------------------------------: | :----------------------------------------------------------------------: |
 | **STEP3-VL-10B-Base** | Base | [🤗 Download](https://huggingface.co/stepfun-ai/Step3-VL-10B-Base) | [🤖 Download](https://modelscope.cn/models/stepfun-ai/Step3-VL-10B-Base) |
 | **STEP3-VL-10B**      | Chat |   [🤗 Download](https://huggingface.co/stepfun-ai/Step3-VL-10B)    |   [🤖 Download](https://modelscope.cn/models/stepfun-ai/Step3-VL-10B)    |
+| **STEP3-VL-10B-FP8**  | Quantized |   [🤗 Download](https://huggingface.co/stepfun-ai/Step3-VL-10B-FP8)    |   [🤖 Download](https://modelscope.cn/models/stepfun-ai/Step3-VL-10B-FP8)    |
 
 ## 📊 Performance