Instructions to use katuni4ka/tiny-random-mistral4-text-only with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use katuni4ka/tiny-random-mistral4-text-only with Transformers:

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

pipe = pipeline("image-text-to-text", model="katuni4ka/tiny-random-mistral4-text-only", trust_remote_code=True)

# Load model directly
from transformers import AutoTokenizer, AutoModelForImageTextToText

tokenizer = AutoTokenizer.from_pretrained("katuni4ka/tiny-random-mistral4-text-only", trust_remote_code=True)
model = AutoModelForImageTextToText.from_pretrained("katuni4ka/tiny-random-mistral4-text-only", trust_remote_code=True)

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use katuni4ka/tiny-random-mistral4-text-only with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "katuni4ka/tiny-random-mistral4-text-only"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "katuni4ka/tiny-random-mistral4-text-only",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/katuni4ka/tiny-random-mistral4-text-only

SGLang

How to use katuni4ka/tiny-random-mistral4-text-only 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 "katuni4ka/tiny-random-mistral4-text-only" \
    --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": "katuni4ka/tiny-random-mistral4-text-only",
		"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 "katuni4ka/tiny-random-mistral4-text-only" \
        --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": "katuni4ka/tiny-random-mistral4-text-only",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use katuni4ka/tiny-random-mistral4-text-only with Docker Model Runner:
```
docker model run hf.co/katuni4ka/tiny-random-mistral4-text-only
```

katuni4ka commited on Apr 3

Commit

d8419e7

verified ·

1 Parent(s): 91af5d5

Update modeling_mistral4.py

Browse files

Files changed (1) hide show

modeling_mistral4.py +8 -4

modeling_mistral4.py CHANGED Viewed

@@ -116,9 +116,12 @@ class Mistral4MoE(nn.Module):
         self.config = config
         self.experts = Mistral4NaiveMoe(config)
         self.gate = Mistral4TopkRouter(config)
-        self.shared_experts = Mistral4MLP(
-            config=config, intermediate_size=config.moe_intermediate_size * config.n_shared_experts
-        )
         self.n_routed_experts = config.n_routed_experts
         self.n_group = config.n_group
         self.topk_group = config.topk_group
@@ -155,7 +158,8 @@ class Mistral4MoE(nn.Module):
         topk_indices, topk_weights = self.route_tokens_to_experts(router_logits)
         hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
         hidden_states = self.experts(hidden_states, topk_indices, topk_weights).view(*orig_shape)
-        hidden_states = hidden_states + self.shared_experts(residuals)
         return hidden_states

         self.config = config
         self.experts = Mistral4NaiveMoe(config)
         self.gate = Mistral4TopkRouter(config)
+        if config.n_shared_experts > 0:
+            self.shared_experts = Mistral4MLP(
+                config=config, intermediate_size=config.moe_intermediate_size * config.n_shared_experts
+            )
+        else:
+            self.shared_experts = None
         self.n_routed_experts = config.n_routed_experts
         self.n_group = config.n_group
         self.topk_group = config.topk_group
         topk_indices, topk_weights = self.route_tokens_to_experts(router_logits)
         hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
         hidden_states = self.experts(hidden_states, topk_indices, topk_weights).view(*orig_shape)
+        if self.shared_experts is not None:
+            hidden_states = hidden_states + self.shared_experts(residuals)
         return hidden_states