Text Generation
Transformers
Safetensors
Vietnamese
English
afmoe
Mixture of Experts
mixture-of-experts
decode-series
llm
vietnamese-llm
Instructions to use Minh2508/Decode with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Minh2508/Decode with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Minh2508/Decode")# Load model directly from transformers import AutoTokenizer, MOE tokenizer = AutoTokenizer.from_pretrained("Minh2508/Decode") model = MOE.from_pretrained("Minh2508/Decode") - Notebooks
- Google Colab
- Kaggle
- Local Apps
- vLLM
How to use Minh2508/Decode with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Minh2508/Decode" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Minh2508/Decode", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Minh2508/Decode
- SGLang
How to use Minh2508/Decode 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 "Minh2508/Decode" \ --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": "Minh2508/Decode", "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 "Minh2508/Decode" \ --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": "Minh2508/Decode", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use Minh2508/Decode with Docker Model Runner:
docker model run hf.co/Minh2508/Decode
Update config.json
Browse files- config.json +38 -38
config.json
CHANGED
|
@@ -1,39 +1,39 @@
|
|
| 1 |
-
{
|
| 2 |
-
"model_type": "
|
| 3 |
-
"architectures": [
|
| 4 |
-
"
|
| 5 |
-
],
|
| 6 |
-
"profile_name": "1b_3e_8l_t4x2",
|
| 7 |
-
"vocab_size": 200024,
|
| 8 |
-
"text_embed_dim": 1024,
|
| 9 |
-
"vision_embed_dim": 1024,
|
| 10 |
-
"hidden_dim": 1024,
|
| 11 |
-
"ffn_dim": 6144,
|
| 12 |
-
"num_layers": 8,
|
| 13 |
-
"num_heads": 16,
|
| 14 |
-
"num_kv_heads": 4,
|
| 15 |
-
"num_experts": 3,
|
| 16 |
-
"top_k": 2,
|
| 17 |
-
"max_position_embeddings": 16384,
|
| 18 |
-
"router_aux_loss_coef": 0.01,
|
| 19 |
-
"share_experts_across_layers": false,
|
| 20 |
-
"gradient_checkpointing": true,
|
| 21 |
-
"num_agents": 4,
|
| 22 |
-
"moe_capacity_factor": 1.0,
|
| 23 |
-
"moe_hierarchy_groups": 1,
|
| 24 |
-
"moe_hierarchy_top_k": 1,
|
| 25 |
-
"num_shared_experts": 0,
|
| 26 |
-
"load_balancing_mode": "aux_free",
|
| 27 |
-
"router_bias_update_rate": 0.01,
|
| 28 |
-
"kv_latent_dim": 128,
|
| 29 |
-
"kv_cache_dtype": "int4",
|
| 30 |
-
"rope_training_context": 16384,
|
| 31 |
-
"rope_ntk_alpha": 1.0,
|
| 32 |
-
"rope_yarn_scale": 1.0,
|
| 33 |
-
"ring_attention_chunk_size": 0,
|
| 34 |
-
"prefill_chunk_size": 256,
|
| 35 |
-
"use_q_former_projector": true,
|
| 36 |
-
"q_former_queries": 8,
|
| 37 |
-
"q_former_layers": 1,
|
| 38 |
-
"tokenizer_name": "ai-tokenizer:GPT-5"
|
| 39 |
}
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "afmoe",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"MOE"
|
| 5 |
+
],
|
| 6 |
+
"profile_name": "1b_3e_8l_t4x2",
|
| 7 |
+
"vocab_size": 200024,
|
| 8 |
+
"text_embed_dim": 1024,
|
| 9 |
+
"vision_embed_dim": 1024,
|
| 10 |
+
"hidden_dim": 1024,
|
| 11 |
+
"ffn_dim": 6144,
|
| 12 |
+
"num_layers": 8,
|
| 13 |
+
"num_heads": 16,
|
| 14 |
+
"num_kv_heads": 4,
|
| 15 |
+
"num_experts": 3,
|
| 16 |
+
"top_k": 2,
|
| 17 |
+
"max_position_embeddings": 16384,
|
| 18 |
+
"router_aux_loss_coef": 0.01,
|
| 19 |
+
"share_experts_across_layers": false,
|
| 20 |
+
"gradient_checkpointing": true,
|
| 21 |
+
"num_agents": 4,
|
| 22 |
+
"moe_capacity_factor": 1.0,
|
| 23 |
+
"moe_hierarchy_groups": 1,
|
| 24 |
+
"moe_hierarchy_top_k": 1,
|
| 25 |
+
"num_shared_experts": 0,
|
| 26 |
+
"load_balancing_mode": "aux_free",
|
| 27 |
+
"router_bias_update_rate": 0.01,
|
| 28 |
+
"kv_latent_dim": 128,
|
| 29 |
+
"kv_cache_dtype": "int4",
|
| 30 |
+
"rope_training_context": 16384,
|
| 31 |
+
"rope_ntk_alpha": 1.0,
|
| 32 |
+
"rope_yarn_scale": 1.0,
|
| 33 |
+
"ring_attention_chunk_size": 0,
|
| 34 |
+
"prefill_chunk_size": 256,
|
| 35 |
+
"use_q_former_projector": true,
|
| 36 |
+
"q_former_queries": 8,
|
| 37 |
+
"q_former_layers": 1,
|
| 38 |
+
"tokenizer_name": "ai-tokenizer:GPT-5"
|
| 39 |
}
|