Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,35 @@
+---
+license: mit
+base_model: MiniMaxAI/MiniMax-M3
+tags:
+- eagle3
+- speculative-decoding
+- compressed-tensors
+- int4
+---
+
+# MiniMax-M3 EAGLE3 drafter — RTN INT4
+
+EAGLE3 speculative-decoding draft head for
+[Sebesky/MiniMax-M3-W4A16-GPTQ](https://huggingface.co/Sebesky/MiniMax-M3-W4A16-GPTQ):
+a single Llama-style decoder layer (hidden 6144, 64-head full MHA, head_dim 128)
+with its **own** `embed_tokens` and `lm_head` (vocab 200,064) — not shared with
+the target. Everything is RTN-quantized to int4 (compressed-tensors
+`pack-quantized`, group 128): Linear, embedding, and lm_head. ~1.6 GB.
+
+## Usage (vLLM)
+
+```
+--speculative-config '{"method":"eagle3","model":"Sebesky/MiniMax-M3-EAGLE3-RTN-INT4","num_speculative_tokens":3}'
+```
+
+Validated on 2x DGX Spark (GB10) at TP2 with `draft_tensor_parallel_size: 2`:
+acceptance length 3.36 (78.6% draft acceptance) at 120k-token context,
+decode 39.7 / 30.4 / 26.4 tok/s @512/65k/120k with 4-bit KV quantization.
+
+Reference deployment (container image, recipes, GB10 operational notes):
+see the `DEPLOYMENT.md` of the companion spark-vllm-docker fork.
+
+Note: the int4 embedding requires the draft model's quantization config to be
+applied to `embed_tokens` at load; vLLM builds that need the
+`fix-eagle3-draft-embed-quant` mod from the deployment repo.