Instructions to use trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration with Transformers:

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

pipe = pipeline("image-text-to-text", model="trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration")
model = AutoModelForMultimodalLM.from_pretrained("trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration

SGLang

How to use trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration 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 "trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration" \
    --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": "trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration" \
        --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": "trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration with Docker Model Runner:
```
docker model run hf.co/trl-internal-testing/tiny-MistralLarge4ForConditionalGeneration
```

qgallouedec HF Staff commited on 3 days ago

Commit

db367c5

verified ·

1 Parent(s): 500a6dd

FP8 weights, like the reference checkpoint

Browse files

Files changed (3) hide show

config.json +24 -1
model.safetensors +2 -2
tokenizer_config.json +2 -2

config.json CHANGED Viewed

@@ -7,11 +7,33 @@
   "model_type": "mistral_large4",
   "multimodal_projector_bias": false,
   "projector_hidden_act": "gelu",
   "spatial_merge_size": 2,
   "text_config": {
     "attention_bias": false,
     "attention_dropout": 0.0,
     "bos_token_id": 1,
     "eos_token_id": 2,
     "final_logit_softcapping": 30.0,
     "first_k_dense_replace": 1,
@@ -66,7 +88,7 @@
       "sparse"
     ],
     "model_type": "mistral_large4_text",
-    "moe_intermediate_size": 16,
     "n_group": 1,
     "n_routed_experts": 8,
     "n_shared_experts": 1,
@@ -111,6 +133,7 @@
   "transformers_version": "5.21.0.dev0",
   "vision_config": {
     "attention_dropout": 0.0,
     "head_dim": 4,
     "hidden_act": "silu",
     "hidden_size": 16,

   "model_type": "mistral_large4",
   "multimodal_projector_bias": false,
   "projector_hidden_act": "gelu",
+  "quantization_config": {
+    "activation_scheme": "dynamic",
+    "dequantize": false,
+    "expert_dtype": "fp8",
+    "modules_to_convert": null,
+    "modules_to_not_convert": [
+      "model.vision_tower",
+      "model.multi_modal_projector",
+      "lm_head",
+      "kv_b_proj",
+      "indexer.wk",
+      "indexer.weights_proj",
+      "kv_a_proj_with_mqa"
+    ],
+    "quant_method": "fp8",
+    "scale_fmt": "float",
+    "weight_block_size": [
+      32,
+      32
+    ]
+  },
   "spatial_merge_size": 2,
   "text_config": {
     "attention_bias": false,
     "attention_dropout": 0.0,
     "bos_token_id": 1,
+    "dtype": "bfloat16",
     "eos_token_id": 2,
     "final_logit_softcapping": 30.0,
     "first_k_dense_replace": 1,
       "sparse"
     ],
     "model_type": "mistral_large4_text",
+    "moe_intermediate_size": 64,
     "n_group": 1,
     "n_routed_experts": 8,
     "n_shared_experts": 1,
   "transformers_version": "5.21.0.dev0",
   "vision_config": {
     "attention_dropout": 0.0,
+    "dtype": "bfloat16",
     "head_dim": 4,
     "hidden_act": "silu",
     "hidden_size": 16,

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:288001a0565e8174b42f35e6049f7d61d7a9249894fac1c6d8a0fd070375f717
-size 26585376

 version https://git-lfs.github.com/spec/v1
+oid sha256:30e7f2faab9479609a7075553e61bcd2239513e0a6fb7b6b506549154217adbb
+size 26755208

tokenizer_config.json CHANGED Viewed

@@ -2,8 +2,8 @@
   "backend": "tokenizers",
   "bos_token": "<s>",
   "eos_token": "</s>",
-  "is_local": false,
-  "local_files_only": true,
   "model_max_length": 1000000000000000019884624838656,
   "pad_token": "<pad>",
   "processor_class": "PixtralProcessor",

   "backend": "tokenizers",
   "bos_token": "<s>",
   "eos_token": "</s>",
+  "is_local": true,
+  "local_files_only": false,
   "model_max_length": 1000000000000000019884624838656,
   "pad_token": "<pad>",
   "processor_class": "PixtralProcessor",