Instructions to use jonathanjordan21/mos-mamba-6x130m-hf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jonathanjordan21/mos-mamba-6x130m-hf with Transformers:

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

pipe = pipeline("text-generation", model="jonathanjordan21/mos-mamba-6x130m-hf", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("jonathanjordan21/mos-mamba-6x130m-hf", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use jonathanjordan21/mos-mamba-6x130m-hf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "jonathanjordan21/mos-mamba-6x130m-hf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jonathanjordan21/mos-mamba-6x130m-hf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/jonathanjordan21/mos-mamba-6x130m-hf

SGLang

How to use jonathanjordan21/mos-mamba-6x130m-hf 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 "jonathanjordan21/mos-mamba-6x130m-hf" \
    --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": "jonathanjordan21/mos-mamba-6x130m-hf",
		"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 "jonathanjordan21/mos-mamba-6x130m-hf" \
        --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": "jonathanjordan21/mos-mamba-6x130m-hf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use jonathanjordan21/mos-mamba-6x130m-hf with Docker Model Runner:
```
docker model run hf.co/jonathanjordan21/mos-mamba-6x130m-hf
```

jonathanjordan21 commited on Aug 30, 2024

Commit

0dd1cff

verified ·

1 Parent(s): c2627ca

Update modeling_mos_mamba.py

Browse files

Files changed (1) hide show

modeling_mos_mamba.py +6 -6

modeling_mos_mamba.py CHANGED Viewed

@@ -604,12 +604,12 @@ class MoSMambaMixer(nn.Module):
 #                 expert_layer.grad = torch.zeros_like(expert_layer.weight)
 #             current_hidden_states = expert_layer(current_state)
-                current_hidden_states = current_hidden_states.reshape(-1, hidden_dim)
-    #             print(current_hidden_states.shape, final_hidden_states.shape)
-                # However `index_add_` only support torch tensors for indexing so we'll use
-                # the `top_x` tensor here.
-                final_hidden_states.index_add_(0, top_x, current_hidden_states.to(hidden_states.dtype))
         final_hidden_states = final_hidden_states.reshape(batch_size, sequence_length, hidden_dim)
         return final_hidden_states, router_logits

 #                 expert_layer.grad = torch.zeros_like(expert_layer.weight)
 #             current_hidden_states = expert_layer(current_state)
+            current_hidden_states = current_hidden_states.reshape(-1, hidden_dim)
+#             print(current_hidden_states.shape, final_hidden_states.shape)
+            # However `index_add_` only support torch tensors for indexing so we'll use
+            # the `top_x` tensor here.
+            final_hidden_states.index_add_(0, top_x, current_hidden_states.to(hidden_states.dtype))
         final_hidden_states = final_hidden_states.reshape(batch_size, sequence_length, hidden_dim)
         return final_hidden_states, router_logits