Instructions to use RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8")
model = AutoModelForCausalLM.from_pretrained("RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8

SGLang

How to use RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8 with Docker Model Runner:
```
docker model run hf.co/RedHatAI/DeepSeek-V4-Flash-NVFP4-FP8
```

kylesayrs commited on 7 days ago

Commit

e4db372

verified ·

1 Parent(s): 224bbad

Upload folder using huggingface_hub

Browse files

Files changed (1) hide show

config.json +25 -25

config.json CHANGED Viewed

@@ -119,8 +119,8 @@
         },
         "output_activations": null,
         "targets": [
-          "re:model.*attn.*(wkv|wo_a|wo_b|wq_a|wq_b)$",
-          "re:model.*attn\\.compressor.*(wgate|wkv)$"
         ],
         "weights": {
           "actorder": null,
@@ -158,7 +158,7 @@
         },
         "output_activations": null,
         "targets": [
-          "re:model.*mlp.*(gate|up|down)_proj$"
         ],
         "weights": {
           "actorder": null,
@@ -179,27 +179,27 @@
     "format": "mixed-precision",
     "global_compression_ratio": null,
     "ignore": [
-      "model.layers.2.self_attn.compressor.indexer.weights_proj",
-      "model.layers.4.self_attn.compressor.indexer.weights_proj",
-      "model.layers.6.self_attn.compressor.indexer.weights_proj",
-      "model.layers.8.self_attn.compressor.indexer.weights_proj",
-      "model.layers.10.self_attn.compressor.indexer.weights_proj",
-      "model.layers.12.self_attn.compressor.indexer.weights_proj",
-      "model.layers.14.self_attn.compressor.indexer.weights_proj",
-      "model.layers.16.self_attn.compressor.indexer.weights_proj",
-      "model.layers.18.self_attn.compressor.indexer.weights_proj",
-      "model.layers.20.self_attn.compressor.indexer.weights_proj",
-      "model.layers.22.self_attn.compressor.indexer.weights_proj",
-      "model.layers.24.self_attn.compressor.indexer.weights_proj",
-      "model.layers.26.self_attn.compressor.indexer.weights_proj",
-      "model.layers.28.self_attn.compressor.indexer.weights_proj",
-      "model.layers.30.self_attn.compressor.indexer.weights_proj",
-      "model.layers.32.self_attn.compressor.indexer.weights_proj",
-      "model.layers.34.self_attn.compressor.indexer.weights_proj",
-      "model.layers.36.self_attn.compressor.indexer.weights_proj",
-      "model.layers.38.self_attn.compressor.indexer.weights_proj",
-      "model.layers.40.self_attn.compressor.indexer.weights_proj",
-      "model.layers.42.self_attn.compressor.indexer.weights_proj",
       "lm_head"
     ],
     "kv_cache_scheme": null,
@@ -235,4 +235,4 @@
   "use_cache": true,
   "v_head_dim": null,
   "vocab_size": 129280
-}

         },
         "output_activations": null,
         "targets": [
+          "re:.*attn.*(wkv|wo_a|wo_b|wq_a|wq_b)$",
+          "re:.*attn\\.compressor.*(wgate|wkv)$"
         ],
         "weights": {
           "actorder": null,
         },
         "output_activations": null,
         "targets": [
+          "re:.*ffn.*(gate|up|down)_proj$"
         ],
         "weights": {
           "actorder": null,
     "format": "mixed-precision",
     "global_compression_ratio": null,
     "ignore": [
+      "layers.2.attn.indexer.weights_proj",
+      "layers.4.attn.indexer.weights_proj",
+      "layers.6.attn.indexer.weights_proj",
+      "layers.8.attn.indexer.weights_proj",
+      "layers.10.attn.indexer.weights_proj",
+      "layers.12.attn.indexer.weights_proj",
+      "layers.14.attn.indexer.weights_proj",
+      "layers.16.attn.indexer.weights_proj",
+      "layers.18.attn.indexer.weights_proj",
+      "layers.20.attn.indexer.weights_proj",
+      "layers.22.attn.indexer.weights_proj",
+      "layers.24.attn.indexer.weights_proj",
+      "layers.26.attn.indexer.weights_proj",
+      "layers.28.attn.indexer.weights_proj",
+      "layers.30.attn.indexer.weights_proj",
+      "layers.32.attn.indexer.weights_proj",
+      "layers.34.attn.indexer.weights_proj",
+      "layers.36.attn.indexer.weights_proj",
+      "layers.38.attn.indexer.weights_proj",
+      "layers.40.attn.indexer.weights_proj",
+      "layers.42.attn.indexer.weights_proj",
       "lm_head"
     ],
     "kv_cache_scheme": null,
   "use_cache": true,
   "v_head_dim": null,
   "vocab_size": 129280
+}