xinference 0.16.3__py3-none-any.whl → 1.2.1__py3-none-any.whl

xinference/model/llm/transformers/cogagent.py

@@ -0,0 +1,272 @@
+# Copyright 2022-2023 XProbe Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import logging
+import re
+import uuid
+from concurrent.futures import ThreadPoolExecutor
+from typing import Dict, Iterator, List, Literal, Optional, Union
+
+import torch
+
+from ....model.utils import select_device
+from ....types import (
+    ChatCompletion,
+    ChatCompletionChunk,
+    CogagentGenerateConfig,
+    CompletionChunk,
+)
+from ..llm_family import LLMFamilyV1, LLMSpecV1
+from ..utils import (
+    _decode_image,
+    generate_chat_completion,
+    generate_completion_chunk,
+    parse_messages,
+)
+from .core import PytorchChatModel
+from .utils import cache_clean
+
+logger = logging.getLogger(__name__)
+
+
+class CogAgentChatModel(PytorchChatModel):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self._torch_type = None
+        self._device = None
+        self._tokenizer = None
+        self._model = None
+        self._platform: Literal["Mac", "WIN", "Mobile"] | None = "Mac"
+        self._format: Literal[
+            "(Answer in Action-Operation-Sensitive format.)",
+            "(Answer in Status-Plan-Action-Operation format.)",
+            "(Answer in Status-Action-Operation-Sensitive format.)",
+            "(Answer in Status-Action-Operation format.)",
+            "(Answer in Action-Operation format.)",
+        ] | None = "(Answer in Action-Operation-Sensitive format.)"
+
+    @classmethod
+    def match(
+        cls, model_family: "LLMFamilyV1", model_spec: "LLMSpecV1", quantization: str
+    ) -> bool:
+        family = model_family.model_family or model_family.model_name
+        if "cogagent" in family.lower():
+            return True
+        return False
+
+    def load(self, **kwargs):
+        from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+
+        device = self._pytorch_model_config.get("device", "auto")
+        self._device = select_device(device)
+
+        self._tokenizer = AutoTokenizer.from_pretrained(
+            self.model_path, trust_remote_code=True
+        )
+        if self.quantization == "4-bit":
+            quantization_config = BitsAndBytesConfig(load_in_4bit=True)
+        elif self.quantization == "8-bit":
+            quantization_config = BitsAndBytesConfig(load_in_8bit=True)
+        else:
+            quantization_config = None
+
+        self._model = AutoModelForCausalLM.from_pretrained(
+            self.model_path,
+            torch_dtype=torch.bfloat16,
+            trust_remote_code=True,
+            device_map=self._device,
+            quantization_config=quantization_config,
+        ).eval()
+
+    def _message_content_to_cogagent(self, content):
+        assert isinstance(content, list)
+        texts = []
+        image_urls = []
+        for c in content:
+            c_type = c.get("type")
+            if c_type == "text":
+                texts.append(c["text"])
+            elif c_type == "image_url":
+                image_urls.append(c["image_url"]["url"])
+        image_futures = []
+        with ThreadPoolExecutor() as executor:
+            for image_url in image_urls:
+                fut = executor.submit(_decode_image, image_url)
+                image_futures.append(fut)
+        images = [fut.result() for fut in image_futures]
+        text = " ".join(texts)
+        if len(images) == 0:
+            raise RuntimeError(
+                "CogAgent requires image input to perform GUI Agent tasks. Pure text-based interaction cannot execute such tasks."
+            )
+        elif len(images) == 1:
+            return text, images[-1]
+        else:
+            logger.warning(
+                "There are multiple images in the prompt, CogAgent will automatically use the most recently provided image as the input."
+            )
+            return text, images[-1]
+
+    def _history_content_to_cogagent(self, chat_history: List[Dict]):
+        grounded_pattern = r"Grounded Operation:\s*(.*)"
+        action_pattern = r"Action:\s*(.*)"
+
+        def extract_operations(_content: str):
+            """extract grounded operation and action operation"""
+            _history_step = []
+            _history_action = []
+
+            matches_history = re.search(grounded_pattern, _content)
+            matches_actions = re.search(action_pattern, _content)
+
+            if matches_history:
+                grounded_operation = matches_history.group(1)
+                _history_step.append(grounded_operation)
+            if matches_actions:
+                action_operation = matches_actions.group(1)
+                _history_action.append(action_operation)
+
+            return _history_step, _history_action
+
+        history_step = []
+        history_action = []
+
+        for i in range(0, len(chat_history) - 1, 2):
+            content = chat_history[i + 1].get("content")
+            if isinstance(content, str):  # 如果内容是字符串
+                steps, actions = extract_operations(content)
+                history_step.extend(steps)
+                history_action.extend(actions)
+
+            elif isinstance(content, list):  # 如果内容是列表
+                for c in content:
+                    c_content = c.get("content")
+                    if isinstance(c_content, str):  # 确保是字符串类型
+                        steps, actions = extract_operations(c_content)
+                        history_step.extend(steps)
+                        history_action.extend(actions)
+
+        return history_step, history_action
+
+    def get_query_and_history(
+        self,
+        prompt: Union[str, List[Dict]],
+        chat_history: Optional[List[Dict]] = None,
+    ):
+        task, image = self._message_content_to_cogagent(prompt)
+
+        history_step, history_action = [], []
+
+        if chat_history:
+            history_step, history_action = self._history_content_to_cogagent(
+                chat_history
+            )
+
+        # Verify history lengths match
+        if len(history_step) != len(history_action):
+            raise ValueError("Mismatch in lengths of history_step and history_action.")
+
+        # Format history steps for output
+        history_str = "\nHistory steps: "
+        for index, (step, action) in enumerate(zip(history_step, history_action)):
+            history_str += f"\n{index}. {step}\t{action}"
+
+        # Compose the query with task, platform, and selected format instructions
+        query = f"Task: {task}{history_str}\n{self._platform}{self._format}"
+        logger.info(f"query:{query}")
+        return query, image
+
+    @cache_clean
+    def chat(
+        self,
+        messages: List[Dict],
+        generate_config: Optional[CogagentGenerateConfig] = None,
+    ) -> Union[ChatCompletion, Iterator[ChatCompletionChunk]]:
+        if generate_config is not None:
+            self._platform = generate_config.pop("platform", self._platform)
+            self._format = generate_config.pop("format", self._format)
+
+        sanitize_generate_config = self._sanitize_generate_config(generate_config)
+        stream = sanitize_generate_config.get("stream")
+        sanitized_config = {
+            "max_length": sanitize_generate_config.get("max_tokens", 512),
+            "top_k": sanitize_generate_config.get("top_k", 1),
+            "do_sample": True,
+        }
+        prompt, _, chat_history = parse_messages(messages)
+
+        query, image = self.get_query_and_history(prompt, chat_history)
+
+        full_context_kwargs = {
+            "return_tensors": "pt",
+            "return_dict": True,
+        }
+        assert self.model_family.chat_template is not None
+        inputs = self.get_full_context(
+            [{"role": "user", "image": image, "content": query}],
+            self.model_family.chat_template,
+            self._tokenizer,
+            tokenize=True,
+            **full_context_kwargs,
+        )
+        inputs.to(self._model.device)
+
+        if stream:
+            it = self._streaming_chat_response(inputs, sanitized_config)
+            return self._to_chat_completion_chunks(it)
+        else:
+            # Generate response
+            with torch.no_grad():
+                outputs = self._model.generate(**inputs, **sanitized_config)
+                outputs = outputs[:, inputs["input_ids"].shape[1] :]
+                response = self._tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+            return generate_chat_completion(self.model_uid, response)
+
+    def _streaming_chat_response(
+        self, inputs: Dict, config: Dict
+    ) -> Iterator[CompletionChunk]:
+        from threading import Thread
+
+        from transformers import TextIteratorStreamer
+
+        streamer = TextIteratorStreamer(
+            self._tokenizer, skip_prompt=True, skip_special_tokens=True
+        )
+        generation_kwargs = {**inputs, **config}
+
+        thread = Thread(target=self._model.generate, kwargs=generation_kwargs)
+        thread.start()
+
+        completion_id = str(uuid.uuid1())
+        for new_text in streamer:
+            yield generate_completion_chunk(
+                chunk_text=new_text,
+                finish_reason=None,
+                chunk_id=completion_id,
+                model_uid=self.model_uid,
+                prompt_tokens=-1,
+                completion_tokens=-1,
+                total_tokens=-1,
+            )
+        yield generate_completion_chunk(
+            chunk_text=None,
+            finish_reason="stop",
+            chunk_id=completion_id,
+            model_uid=self.model_uid,
+            prompt_tokens=-1,
+            completion_tokens=-1,
+            total_tokens=-1,
+            has_choice=True,
+            has_content=False,
+        )

xinference/model/llm/transformers/core.py

@@ -68,6 +68,9 @@ NON_DEFAULT_MODEL_LIST: List[str] = [
     "deepseek-v2-chat",
     "deepseek-v2.5",
     "deepseek-v2-chat-0628",
+    "glm-edge-v",
+    "QvQ-72B-Preview",
+    "cogagent",
 ]
 
 

xinference/model/llm/transformers/glm_edge_v.py

@@ -0,0 +1,230 @@
+# Copyright 2022-2023 XProbe Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import logging
+import uuid
+from concurrent.futures import ThreadPoolExecutor
+from threading import Thread
+from typing import Any, Dict, Iterator, List, Optional, Tuple, Union
+
+import torch
+
+from ....types import ChatCompletion, ChatCompletionChunk, CompletionChunk
+from ...utils import select_device
+from ..llm_family import LLMFamilyV1, LLMSpecV1
+from ..utils import (
+    _decode_image_without_rgb,
+    generate_chat_completion,
+    generate_completion_chunk,
+)
+from .core import PytorchChatModel, PytorchGenerateConfig
+from .utils import cache_clean
+
+logger = logging.getLogger(__name__)
+
+
+class GlmEdgeVModel(PytorchChatModel):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self._device = None
+        self._tokenizer = None
+        self._model = None
+        self._processor = None
+
+    @classmethod
+    def match(
+        cls, model_family: "LLMFamilyV1", model_spec: "LLMSpecV1", quantization: str
+    ) -> bool:
+        family = model_family.model_family or model_family.model_name
+        if "glm-edge-v" in family.lower():
+            return True
+        return False
+
+    def load(self):
+        from transformers import AutoImageProcessor, AutoModelForCausalLM, AutoTokenizer
+
+        device = self._pytorch_model_config.get("device", "auto")
+        self._device = select_device(device)
+
+        kwargs = {"device_map": self._device}
+        quantization = self.quantization
+
+        # referenced from PytorchModel.load
+        if quantization != "none":
+            if self._device == "cuda" and self._is_linux():
+                kwargs["device_map"] = "auto"
+                if quantization == "4-bit":
+                    kwargs["load_in_4bit"] = True
+                elif quantization == "8-bit":
+                    kwargs["load_in_8bit"] = True
+                else:
+                    raise ValueError(
+                        f"Quantization {quantization} is not supported in temporary"
+                    )
+            else:
+                if quantization != "8-bit":
+                    raise ValueError(
+                        f"Only 8-bit quantization is supported if it is not linux system or cuda device"
+                    )
+
+        processor = AutoImageProcessor.from_pretrained(
+            self.model_path, trust_remote_code=True
+        )
+        self._processor = processor
+
+        model = AutoModelForCausalLM.from_pretrained(
+            self.model_path,
+            trust_remote_code=True,
+            torch_dtype=torch.bfloat16,
+            device_map="auto",
+        )
+
+        self._model = model
+
+        tokenizer = AutoTokenizer.from_pretrained(
+            self.model_path, trust_remote_code=True
+        )
+        self._tokenizer = tokenizer
+
+    @staticmethod
+    def _get_processed_msgs(
+        messages: List[Dict],
+    ) -> Tuple[List[Dict[str, Any]], List[Any]]:
+        res = []
+        img = []
+        for message in messages:
+            role = message["role"]
+            content = message["content"]
+            if isinstance(content, str):
+                res.append({"role": role, "content": content})
+            else:
+                texts = []
+                image_urls = []
+                for c in content:
+                    c_type = c.get("type")
+                    if c_type == "text":
+                        texts.append(c["text"])
+                    else:
+                        assert (
+                            c_type == "image_url"
+                        ), "Please follow the image input of the OpenAI API."
+                        image_urls.append(c["image_url"]["url"])
+                if len(image_urls) > 1:
+                    raise RuntimeError("Only one image per message is supported")
+                image_futures = []
+                with ThreadPoolExecutor() as executor:
+                    for image_url in image_urls:
+                        fut = executor.submit(_decode_image_without_rgb, image_url)
+                        image_futures.append(fut)
+                images = [fut.result() for fut in image_futures]
+                assert len(images) <= 1
+                text = " ".join(texts)
+                img.extend(images)
+                if images:
+                    res.append(
+                        {
+                            "role": role,
+                            "content": [
+                                {"type": "image"},
+                                {"type": "text", "text": text},
+                            ],
+                        }
+                    )
+                else:
+                    res.append({"role": role, "content": text})
+        return res, img
+
+    @cache_clean
+    def chat(
+        self,
+        messages: List[Dict],
+        generate_config: Optional[PytorchGenerateConfig] = None,
+    ) -> Union[ChatCompletion, Iterator[ChatCompletionChunk]]:
+        from transformers import TextIteratorStreamer
+
+        if not generate_config:
+            generate_config = {}
+
+        stream = generate_config.get("stream", False)
+        msgs, imgs = self._get_processed_msgs(messages)
+
+        inputs = self._tokenizer.apply_chat_template(
+            msgs,
+            add_generation_prompt=True,
+            tokenize=True,
+            return_tensors="pt",
+            return_dict=True,
+        )  # chat mode
+        inputs = inputs.to(self._model.device)
+
+        generate_kwargs = {
+            **inputs,
+        }
+        if len(imgs) > 0:
+            generate_kwargs["pixel_values"] = torch.tensor(
+                self._processor(imgs[-1]).pixel_values
+            ).to(self._model.device)
+        stop_str = "<|endoftext|>"
+
+        if stream:
+            streamer = TextIteratorStreamer(
+                tokenizer=self._tokenizer,
+                timeout=60,
+                skip_prompt=True,
+                skip_special_tokens=True,
+            )
+            generate_kwargs = {
+                **generate_kwargs,
+                "streamer": streamer,
+            }
+            t = Thread(target=self._model.generate, kwargs=generate_kwargs)
+            t.start()
+
+            it = self.chat_stream(streamer, stop_str)
+            return self._to_chat_completion_chunks(it)
+        else:
+            with torch.no_grad():
+                outputs = self._model.generate(**generate_kwargs)
+                outputs = outputs[0][len(inputs["input_ids"][0]) :]
+                response = self._tokenizer.decode(outputs)
+                if response.endswith(stop_str):
+                    response = response[: -len(stop_str)]
+            return generate_chat_completion(self.model_uid, response)
+
+    def chat_stream(self, streamer, stop_str) -> Iterator[CompletionChunk]:
+        completion_id = str(uuid.uuid1())
+        for new_text in streamer:
+            if not new_text.endswith(stop_str):
+                yield generate_completion_chunk(
+                    chunk_text=new_text,
+                    finish_reason=None,
+                    chunk_id=completion_id,
+                    model_uid=self.model_uid,
+                    prompt_tokens=-1,
+                    completion_tokens=-1,
+                    total_tokens=-1,
+                    has_choice=True,
+                    has_content=True,
+                )
+
+        yield generate_completion_chunk(
+            chunk_text=None,
+            finish_reason="stop",
+            chunk_id=completion_id,
+            model_uid=self.model_uid,
+            prompt_tokens=-1,
+            completion_tokens=-1,
+            total_tokens=-1,
+            has_choice=True,
+            has_content=False,
+        )

xinference/model/llm/transformers/qwen2_vl.py

@@ -17,6 +17,7 @@ import sys
 import uuid
 from typing import Iterator, List, Optional, Union
 
+from ....device_utils import is_npu_available
 from ....model.utils import select_device
 from ....types import (
     ChatCompletion,
@@ -47,6 +48,8 @@ class Qwen2VLChatModel(PytorchChatModel):
         llm_family = model_family.model_family or model_family.model_name
         if "qwen2-vl-instruct".lower() in llm_family.lower():
             return True
+        if "qvq-72b-preview".lower() in llm_family.lower():
+            return True
         return False
 
     def load(self):
@@ -71,6 +74,14 @@ class Qwen2VLChatModel(PytorchChatModel):
                 attn_implementation="flash_attention_2",
                 trust_remote_code=True,
             ).eval()
+        elif is_npu_available():
+            # Ascend do not support bf16
+            self._model = Qwen2VLForConditionalGeneration.from_pretrained(
+                self.model_path,
+                device_map="auto",
+                trust_remote_code=True,
+                torch_dtype="float16",
+            ).eval()
         else:
             self._model = Qwen2VLForConditionalGeneration.from_pretrained(
                 self.model_path, device_map=device, trust_remote_code=True
@@ -112,7 +123,7 @@ class Qwen2VLChatModel(PytorchChatModel):
             padding=True,
             return_tensors="pt",
         )
-        inputs = inputs.to("cuda")
+        inputs = inputs.to(self._device)
 
         # Inference: Generation of the output
         generated_ids = self._model.generate(

xinference/model/llm/transformers/utils.py

@@ -156,6 +156,7 @@ def _get_completion(
     finish_reason: Optional[str],
     model_uid: str,
     r: InferenceRequest,
+    completion_tokens: int,
 ):
     completion_choice = CompletionChoice(
         text=output, index=0, logprobs=None, finish_reason=finish_reason
@@ -170,8 +171,8 @@ def _get_completion(
     )
     completion_usage = CompletionUsage(
         prompt_tokens=len(r.prompt_tokens),
-        completion_tokens=len(r.new_tokens),
-        total_tokens=len(r.prompt_tokens) + len(r.new_tokens),
+        completion_tokens=completion_tokens,
+        total_tokens=len(r.prompt_tokens) + completion_tokens,
     )
     completion = Completion(
         id=completion_chunk["id"],
@@ -371,7 +372,7 @@ def _batch_inference_one_step_internal(
                 r.stopped = stopped
                 r.finish_reason = finish_reason
 
-            if r.stopped and r not in stop_token_mapping and r not in output_mapping:
+            if r.stopped and r not in stop_token_mapping:
                 stop_token_mapping[r] = _i + 1
 
             if r.stream:
@@ -446,12 +447,14 @@ def _batch_inference_one_step_internal(
             else:
                 # last round, handle non-stream result
                 if r.stopped and _i == decode_round - 1:
-                    invalid_token_num = decode_round - stop_token_mapping[r]
+                    invalid_token_num = (
+                        (decode_round - stop_token_mapping[r] + 1)
+                        if r.finish_reason == "stop"
+                        else (decode_round - stop_token_mapping[r])
+                    )
                     outputs = (
                         tokenizer.decode(
-                            r.new_tokens[: -(invalid_token_num + 1)]
-                            if r.finish_reason == "stop"
-                            else r.new_tokens[:-invalid_token_num],
+                            r.new_tokens[:-invalid_token_num],
                             skip_special_tokens=True,
                             spaces_between_special_tokens=False,
                             clean_up_tokenization_spaces=True,
@@ -460,7 +463,12 @@ def _batch_inference_one_step_internal(
                         else output_mapping[r]
                     )
                     completion = _get_completion(
-                        outputs, r.chunk_id, r.finish_reason, model_uid, r
+                        outputs,
+                        r.chunk_id,
+                        r.finish_reason,
+                        model_uid,
+                        r,
+                        len(r.new_tokens) - invalid_token_num,
                     )
                     r.completion = [completion]