Instructions to use silx-ai/Quasar-3B-A1B-Preview with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use silx-ai/Quasar-3B-A1B-Preview with Transformers:

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

pipe = pipeline("text-generation", model="silx-ai/Quasar-3B-A1B-Preview", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("silx-ai/Quasar-3B-A1B-Preview", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use silx-ai/Quasar-3B-A1B-Preview with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "silx-ai/Quasar-3B-A1B-Preview"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "silx-ai/Quasar-3B-A1B-Preview",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/silx-ai/Quasar-3B-A1B-Preview

SGLang

How to use silx-ai/Quasar-3B-A1B-Preview 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 "silx-ai/Quasar-3B-A1B-Preview" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "silx-ai/Quasar-3B-A1B-Preview",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "silx-ai/Quasar-3B-A1B-Preview" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "silx-ai/Quasar-3B-A1B-Preview",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use silx-ai/Quasar-3B-A1B-Preview with Docker Model Runner:
```
docker model run hf.co/silx-ai/Quasar-3B-A1B-Preview
```

eyad-silx commited on Apr 27

Commit

c0e30fb

verified ·

1 Parent(s): ec9d21b

Update config.json

Browse files

Files changed (1) hide show

config.json +44 -15

config.json CHANGED Viewed

@@ -1,24 +1,53 @@
 {
   "model_type": "quasar",
-  "auto_map": {
-    "AutoConfig": "configuration_quasar.QuasarConfig",
-    "AutoModelForCausalLM": "modeling_quasar.QuasarForCausalLM"
-  },
   "vocab_size": 248320,
-  "d_model": 2048,
-  "n_heads": 16,
-  "n_layers": 20,
-  "d_ff": 5632,
-  "num_shared_experts": 1,
-  "num_routed_experts": 48,
-  "top_k": 4,
-  "shared_expert_size": 1536,
-  "routed_expert_size": 512,
   "quasar_layers": 4,
   "gated_layers": 2,
   "dense_input_layers": 4,
-  "max_seq_len": 16384,
   "num_loops": 1,
   "rope_theta": 1000000.0,
-  "hidden_act": "silu"
 }

 {
   "model_type": "quasar",
   "vocab_size": 248320,
+  "d_model": 1536,
+  "n_heads": 12,
+  "n_layers": 24,
+  "d_ff": 4096,
+  "head_dim": 128,
+  "max_seq_len": 16384,
+  "dropout": 0.0,
+  "rms_norm_eps": 1e-6,
+  "initializer_range": 0.02,
+  "use_cache": true,
+  "tie_word_embeddings": false,
   "quasar_layers": 4,
   "gated_layers": 2,
+  "use_gla_first": false,
+  "use_short_conv": true,
+  "conv_size": 4,
+  "conv_bias": false,
+  "allow_neg_eigval": false,
+  "attn_mode": "chunk",
+  "expand_k": 0.5,
+  "expand_v": 1.0,
+  "gla_mode": "chunk",
+  "memory_slots": 128,
+  "memory_dim": 128,
+  "moe_type": "bigmac",
+  "num_shared_experts": 1,
+  "num_routed_experts": 64,
+  "top_k": 4,
+  "shared_expert_size": 3072,
+  "routed_expert_size": 256,
   "dense_input_layers": 4,
+  "bigmac_r": 0.25,
+  "moe_z_loss_coeff": 0.0001,
+  "moe_aux_loss_coeff": 0.0001,
+  "smebu_kappa": 2.0,
+  "smebu_lambda": 0.002,
+  "smebu_beta": 0.5,
   "num_loops": 1,
+  "use_looped_injection": false,
+  "looped_injection_init": 0.1,
   "rope_theta": 1000000.0,
+  "hidden_act": "silu",
+  "residual_scale": 0.1,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "auto_map": {
+    "AutoConfig": "configuration_quasar.QuasarConfig",
+    "AutoModelForCausalLM": "modeling_quasar.QuasarForCausalLM"
+  }
 }