|
|
--- |
|
|
license: mit |
|
|
base_model: |
|
|
- perplexity-ai/r1-1776 |
|
|
--- |
|
|
|
|
|
DO NOT USE. Currently vLLM does not support FP8 Dynamic on FP8 Fused MoE layers. Keeping this model for later use. |
|
|
|
|
|
Run cmd: |
|
|
|
|
|
```bash |
|
|
NB_GPU=8 |
|
|
docker run --rm --runtime nvidia --gpus 'all' --ipc=host -e "OMP_NUM_THREADS=$(nproc)" -e 'HF_TOKEN' -v '/root/.cache/huggingface:/root/.cache/huggingface' -p 8000:8000 'vllm/vllm-openai:v0.7.3' --host 0.0.0.0 --port 8000 --trust-remote-code --tensor-parallel-size $NB_GPU --served-model-name deepseek-reasoner --enable-reasoning --reasoning-parser deepseek_r1 --model 'ig1/r1-1776-FP8-Dynamic' --override-generation-config '{"temperature":0.6,"top_p":0.95}' --enable-chunked-prefill=False |
|
|
``` |
|
|
|
|
|
Error: |
|
|
|
|
|
``` |
|
|
For FP8 Fused MoE layers, only per-tensor scalesfor weights and activations are supported. Found num_bits=8 type='float' symmetric=True group_size=None strategy='channel' block_structure=None dynamic=False actorder=None observer='minmax' observer_kwargs={}, num_bits=8 type='float' symmetric=True group_size=None strategy='token' block_structure=None dynamic=True actorder=None observer=None observer_kwargs={} |
|
|
``` |
