初始化项目，由ModelHub XC社区提供模型
Model: ds4sd/SmolDocling-256M-preview Source: Original Platform
This commit is contained in:
ModelHub XC
72
.gitattributes
vendored
Normal file
72
.gitattributes
vendored
Normal file
@@ -0,0 +1,72 @@
|
|||||||
|
*.7z filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.arrow filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.bin filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.bin.* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.ftz filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.gz filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.h5 filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.joblib filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.model filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.ot filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.parquet filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.pb filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.pt filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.pth filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.rar filter=lfs diff=lfs merge=lfs -text
|
||||||
|
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.tflite filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.tgz filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.xz filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.zip filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.zstandard filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.tfevents* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.db* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.ark* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
**/*ckpt*data* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
**/*ckpt*.meta filter=lfs diff=lfs merge=lfs -text
|
||||||
|
**/*ckpt*.index filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.gguf* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.ggml filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.llamafile* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.pt2 filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.npy filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.npz filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.pickle filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.pkl filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.tar filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.wasm filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.zst filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/decoder_model_merged.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/decoder_model_merged_bnb4.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/decoder_model_merged_fp16.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/decoder_model_merged_int8.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/decoder_model_merged_q4.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/decoder_model_merged_q4f16.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/decoder_model_merged_quantized.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/decoder_model_merged_uint8.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/embed_tokens.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/embed_tokens_bnb4.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/embed_tokens_fp16.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/embed_tokens_int8.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/embed_tokens_q4.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/embed_tokens_q4f16.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/embed_tokens_quantized.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/embed_tokens_uint8.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/vision_encoder.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/vision_encoder_bnb4.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/vision_encoder_fp16.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/vision_encoder_int8.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/vision_encoder_q4.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/vision_encoder_q4f16.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/vision_encoder_quantized.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
onnx/vision_encoder_uint8.onnx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
model.safetensors filter=lfs diff=lfs merge=lfs -text
|
||||||
432
README.md
Normal file
432
README.md
Normal file
@@ -0,0 +1,432 @@
|
|||||||
|
---
|
||||||
|
base_model:
|
||||||
|
- HuggingFaceTB/SmolVLM-256M-Instruct
|
||||||
|
language:
|
||||||
|
- en
|
||||||
|
library_name: transformers
|
||||||
|
license: cdla-permissive-2.0
|
||||||
|
pipeline_tag: image-text-to-text
|
||||||
|
datasets:
|
||||||
|
- ds4sd/SynthCodeNet
|
||||||
|
- ds4sd/SynthFormulaNet
|
||||||
|
- ds4sd/SynthChartNet
|
||||||
|
- HuggingFaceM4/DoclingMatix
|
||||||
|
---
|
||||||
|
|
||||||
|
<div style="
|
||||||
|
background-color: #f0f9ff;
|
||||||
|
border: 1px solid #bae6fd;
|
||||||
|
color: #0369a1;
|
||||||
|
padding: 12px 16px;
|
||||||
|
border-radius: 12px;
|
||||||
|
margin-bottom: 16px;
|
||||||
|
font-family: sans-serif;
|
||||||
|
">
|
||||||
|
<strong>📢 New Release:</strong>
|
||||||
|
We’ve released <a href="https://huggingface.co/ibm-granite/granite-docling-258M" target="_blank" style="color:#0284c7; font-weight:bold; text-decoration:underline;">
|
||||||
|
granite-docling-258M</a>, the successor to <b>SmolDocling</b>. It will now receive updates and support, check it out!
|
||||||
|
</div>
|
||||||
|
|
||||||
|
<div style="display: flex; align-items: center;">
|
||||||
|
<img src="https://huggingface.co/ds4sd/SmolDocling-256M-preview/resolve/main/assets/SmolDocling_doctags1.png" alt="SmolDocling" style="width: 200px; height: auto; margin-right: 20px;">
|
||||||
|
<div>
|
||||||
|
<h3>SmolDocling-256M-preview</h3>
|
||||||
|
<p>SmolDocling is a multimodal Image-Text-to-Text model designed for efficient document conversion. It retains Docling's most popular features while ensuring full compatibility with Docling through seamless support for <strong>DoclingDocuments</strong>.</p>
|
||||||
|
</div>
|
||||||
|
</div>
|
||||||
|
|
||||||
|
This model was presented in the paper [SmolDocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion](https://huggingface.co/papers/2503.11576).
|
||||||
|
|
||||||
|
### 🚀 Features:
|
||||||
|
- 🏷️ **DocTags for Efficient Tokenization** – Introduces DocTags an efficient and minimal representation for documents that is fully compatible with **DoclingDocuments**.
|
||||||
|
- 🔍 **OCR (Optical Character Recognition)** – Extracts text accurately from images.
|
||||||
|
- 📐 **Layout and Localization** – Preserves document structure and document element **bounding boxes**.
|
||||||
|
- 💻 **Code Recognition** – Detects and formats code blocks including identation.
|
||||||
|
- 🔢 **Formula Recognition** – Identifies and processes mathematical expressions.
|
||||||
|
- 📊 **Chart Recognition** – Extracts and interprets chart data.
|
||||||
|
- 📑 **Table Recognition** – Supports column and row headers for structured table extraction.
|
||||||
|
- 🖼️ **Figure Classification** – Differentiates figures and graphical elements.
|
||||||
|
- 📝 **Caption Correspondence** – Links captions to relevant images and figures.
|
||||||
|
- 📜 **List Grouping** – Organizes and structures list elements correctly.
|
||||||
|
- 📄 **Full-Page Conversion** – Processes entire pages for comprehensive document conversion including all page elements (code, equations, tables, charts etc.)
|
||||||
|
- 🔲 **OCR with Bounding Boxes** – OCR regions using a bounding box.
|
||||||
|
- 📂 **General Document Processing** – Trained for both scientific and non-scientific documents.
|
||||||
|
- 🔄 **Seamless Docling Integration** – Import into **Docling** and export in multiple formats.
|
||||||
|
- 💨 **Fast inference using VLLM** – Avg of 0.35 secs per page on A100 GPU.
|
||||||
|
|
||||||
|
### 🚧 *Coming soon!*
|
||||||
|
- 📊 **Better chart recognition 🛠️**
|
||||||
|
- 📚 **One shot multi-page inference ⏱️**
|
||||||
|
- 🧪 **Chemical Recognition**
|
||||||
|
- 📙 **Datasets**
|
||||||
|
|
||||||
|
## ⌨️ Get started (code examples)
|
||||||
|
|
||||||
|
You can use **transformers**, **vllm**, or **onnx** to perform inference, and [Docling](https://github.com/docling-project/docling) to convert results to variety of output formats (md, html, etc.):
|
||||||
|
|
||||||
|
<details>
|
||||||
|
<summary>📄 Single page image inference using Tranformers 🤖</summary>
|
||||||
|
|
||||||
|
```python
|
||||||
|
# Prerequisites:
|
||||||
|
# pip install torch
|
||||||
|
# pip install docling_core
|
||||||
|
# pip install transformers
|
||||||
|
|
||||||
|
import torch
|
||||||
|
from docling_core.types.doc import DoclingDocument
|
||||||
|
from docling_core.types.doc.document import DocTagsDocument
|
||||||
|
from transformers import AutoProcessor, AutoModelForVision2Seq
|
||||||
|
from transformers.image_utils import load_image
|
||||||
|
from pathlib import Path
|
||||||
|
|
||||||
|
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
||||||
|
|
||||||
|
# Load images
|
||||||
|
image = load_image("https://upload.wikimedia.org/wikipedia/commons/7/76/GazettedeFrance.jpg")
|
||||||
|
|
||||||
|
# Initialize processor and model
|
||||||
|
processor = AutoProcessor.from_pretrained("ds4sd/SmolDocling-256M-preview")
|
||||||
|
model = AutoModelForVision2Seq.from_pretrained(
|
||||||
|
"ds4sd/SmolDocling-256M-preview",
|
||||||
|
torch_dtype=torch.bfloat16,
|
||||||
|
_attn_implementation="flash_attention_2" if DEVICE == "cuda" else "eager",
|
||||||
|
).to(DEVICE)
|
||||||
|
|
||||||
|
# Create input messages
|
||||||
|
messages = [
|
||||||
|
{
|
||||||
|
"role": "user",
|
||||||
|
"content": [
|
||||||
|
{"type": "image"},
|
||||||
|
{"type": "text", "text": "Convert this page to docling."}
|
||||||
|
]
|
||||||
|
},
|
||||||
|
]
|
||||||
|
|
||||||
|
# Prepare inputs
|
||||||
|
prompt = processor.apply_chat_template(messages, add_generation_prompt=True)
|
||||||
|
inputs = processor(text=prompt, images=[image], return_tensors="pt")
|
||||||
|
inputs = inputs.to(DEVICE)
|
||||||
|
|
||||||
|
# Generate outputs
|
||||||
|
generated_ids = model.generate(**inputs, max_new_tokens=8192)
|
||||||
|
prompt_length = inputs.input_ids.shape[1]
|
||||||
|
trimmed_generated_ids = generated_ids[:, prompt_length:]
|
||||||
|
doctags = processor.batch_decode(
|
||||||
|
trimmed_generated_ids,
|
||||||
|
skip_special_tokens=False,
|
||||||
|
)[0].lstrip()
|
||||||
|
|
||||||
|
# Populate document
|
||||||
|
doctags_doc = DocTagsDocument.from_doctags_and_image_pairs([doctags], [image])
|
||||||
|
print(doctags)
|
||||||
|
# create a docling document
|
||||||
|
doc = DoclingDocument.load_from_doctags(doctags_doc, document_name="Document")
|
||||||
|
|
||||||
|
# export as any format
|
||||||
|
# HTML
|
||||||
|
# Path("Out/").mkdir(parents=True, exist_ok=True)
|
||||||
|
# output_path_html = Path("Out/") / "example.html"
|
||||||
|
# doc.save_as_html(output_path_html)
|
||||||
|
# MD
|
||||||
|
print(doc.export_to_markdown())
|
||||||
|
```
|
||||||
|
</details>
|
||||||
|
|
||||||
|
|
||||||
|
<details>
|
||||||
|
<summary> 🚀 Fast Batch Inference Using VLLM</summary>
|
||||||
|
|
||||||
|
```python
|
||||||
|
# Prerequisites:
|
||||||
|
# pip install vllm
|
||||||
|
# pip install docling_core
|
||||||
|
# place page images you want to convert into "img/" dir
|
||||||
|
|
||||||
|
import time
|
||||||
|
import os
|
||||||
|
from vllm import LLM, SamplingParams
|
||||||
|
from PIL import Image
|
||||||
|
from docling_core.types.doc import DoclingDocument
|
||||||
|
from docling_core.types.doc.document import DocTagsDocument
|
||||||
|
from pathlib import Path
|
||||||
|
|
||||||
|
# Configuration
|
||||||
|
MODEL_PATH = "ds4sd/SmolDocling-256M-preview"
|
||||||
|
IMAGE_DIR = "img/" # Place your page images here
|
||||||
|
OUTPUT_DIR = "out/"
|
||||||
|
PROMPT_TEXT = "Convert page to Docling."
|
||||||
|
|
||||||
|
# Ensure output directory exists
|
||||||
|
os.makedirs(OUTPUT_DIR, exist_ok=True)
|
||||||
|
|
||||||
|
# Initialize LLM
|
||||||
|
llm = LLM(model=MODEL_PATH, limit_mm_per_prompt={"image": 1})
|
||||||
|
|
||||||
|
sampling_params = SamplingParams(
|
||||||
|
temperature=0.0,
|
||||||
|
max_tokens=8192)
|
||||||
|
|
||||||
|
chat_template = f"<|im_start|>User:<image>{PROMPT_TEXT}<end_of_utterance>
|
||||||
|
Assistant:"
|
||||||
|
|
||||||
|
image_files = sorted([f for f in os.listdir(IMAGE_DIR) if f.lower().endswith((".png", ".jpg", ".jpeg"))])
|
||||||
|
|
||||||
|
start_time = time.time()
|
||||||
|
total_tokens = 0
|
||||||
|
|
||||||
|
for idx, img_file in enumerate(image_files, 1):
|
||||||
|
img_path = os.path.join(IMAGE_DIR, img_file)
|
||||||
|
image = Image.open(img_path).convert("RGB")
|
||||||
|
|
||||||
|
llm_input = {"prompt": chat_template, "multi_modal_data": {"image": image}}
|
||||||
|
output = llm.generate([llm_input], sampling_params=sampling_params)[0]
|
||||||
|
|
||||||
|
doctags = output.outputs[0].text
|
||||||
|
img_fn = os.path.splitext(img_file)[0]
|
||||||
|
output_filename = img_fn + ".dt"
|
||||||
|
output_path = os.path.join(OUTPUT_DIR, output_filename)
|
||||||
|
|
||||||
|
with open(output_path, "w", encoding="utf-8") as f:
|
||||||
|
f.write(doctags)
|
||||||
|
|
||||||
|
# To convert to Docling Document, MD, HTML, etc.:
|
||||||
|
doctags_doc = DocTagsDocument.from_doctags_and_image_pairs([doctags], [image])
|
||||||
|
doc = DoclingDocument.load_from_doctags(doctags_doc, document_name="Document")
|
||||||
|
# export as any format
|
||||||
|
# HTML
|
||||||
|
# output_path_html = Path(OUTPUT_DIR) / f"{img_fn}.html"
|
||||||
|
# doc.save_as_html(output_path_html)
|
||||||
|
# MD
|
||||||
|
output_path_md = Path(OUTPUT_DIR) / f"{img_fn}.md"
|
||||||
|
doc.save_as_markdown(output_path_md)
|
||||||
|
print(f"Total time: {time.time() - start_time:.2f} sec")
|
||||||
|
```
|
||||||
|
</details>
|
||||||
|
<details>
|
||||||
|
<summary> ONNX Inference</summary>
|
||||||
|
|
||||||
|
```python
|
||||||
|
# Prerequisites:
|
||||||
|
# pip install onnxruntime
|
||||||
|
# pip install onnxruntime-gpu
|
||||||
|
from transformers import AutoConfig, AutoProcessor
|
||||||
|
from transformers.image_utils import load_image
|
||||||
|
import onnxruntime
|
||||||
|
import numpy as np
|
||||||
|
import os
|
||||||
|
from docling_core.types.doc import DoclingDocument
|
||||||
|
from docling_core.types.doc.document import DocTagsDocument
|
||||||
|
|
||||||
|
os.environ["OMP_NUM_THREADS"] = "1"
|
||||||
|
# cuda
|
||||||
|
os.environ["ORT_CUDA_USE_MAX_WORKSPACE"] = "1"
|
||||||
|
|
||||||
|
# 1. Load models
|
||||||
|
## Load config and processor
|
||||||
|
model_id = "ds4sd/SmolDocling-256M-preview"
|
||||||
|
config = AutoConfig.from_pretrained(model_id)
|
||||||
|
processor = AutoProcessor.from_pretrained(model_id)
|
||||||
|
|
||||||
|
## Load sessions
|
||||||
|
# !wget https://huggingface.co/ds4sd/SmolDocling-256M-preview/resolve/main/onnx/vision_encoder.onnx
|
||||||
|
# !wget https://huggingface.co/ds4sd/SmolDocling-256M-preview/resolve/main/onnx/embed_tokens.onnx
|
||||||
|
# !wget https://huggingface.co/ds4sd/SmolDocling-256M-preview/resolve/main/onnx/decoder_model_merged.onnx
|
||||||
|
# cpu
|
||||||
|
# vision_session = onnxruntime.InferenceSession("vision_encoder.onnx")
|
||||||
|
# embed_session = onnxruntime.InferenceSession("embed_tokens.onnx")
|
||||||
|
# decoder_session = onnxruntime.InferenceSession("decoder_model_merged.onnx"
|
||||||
|
|
||||||
|
# cuda
|
||||||
|
vision_session = onnxruntime.InferenceSession("vision_encoder.onnx", providers=["CUDAExecutionProvider"])
|
||||||
|
embed_session = onnxruntime.InferenceSession("embed_tokens.onnx", providers=["CUDAExecutionProvider"])
|
||||||
|
decoder_session = onnxruntime.InferenceSession("decoder_model_merged.onnx", providers=["CUDAExecutionProvider"])
|
||||||
|
|
||||||
|
## Set config values
|
||||||
|
num_key_value_heads = config.text_config.num_key_value_heads
|
||||||
|
head_dim = config.text_config.head_dim
|
||||||
|
num_hidden_layers = config.text_config.num_hidden_layers
|
||||||
|
eos_token_id = config.text_config.eos_token_id
|
||||||
|
image_token_id = config.image_token_id
|
||||||
|
end_of_utterance_id = processor.tokenizer.convert_tokens_to_ids("<end_of_utterance>")
|
||||||
|
|
||||||
|
# 2. Prepare inputs
|
||||||
|
## Create input messages
|
||||||
|
messages = [
|
||||||
|
{
|
||||||
|
"role": "user",
|
||||||
|
"content": [
|
||||||
|
{"type": "image"},
|
||||||
|
{"type": "text", "text": "Convert this page to docling."}
|
||||||
|
]
|
||||||
|
},
|
||||||
|
]
|
||||||
|
|
||||||
|
## Load image and apply processor
|
||||||
|
image = load_image("https://ibm.biz/docling-page-with-table")
|
||||||
|
prompt = processor.apply_chat_template(messages, add_generation_prompt=True)
|
||||||
|
inputs = processor(text=prompt, images=[image], return_tensors="np")
|
||||||
|
|
||||||
|
## Prepare decoder inputs
|
||||||
|
batch_size = inputs['input_ids'].shape[0]
|
||||||
|
past_key_values = {
|
||||||
|
f'past_key_values.{layer}.{kv}': np.zeros([batch_size, num_key_value_heads, 0, head_dim], dtype=np.float32)
|
||||||
|
for layer in range(num_hidden_layers)
|
||||||
|
for kv in ('key', 'value')
|
||||||
|
}
|
||||||
|
image_features = None
|
||||||
|
input_ids = inputs['input_ids']
|
||||||
|
attention_mask = inputs['attention_mask']
|
||||||
|
position_ids = np.cumsum(inputs['attention_mask'], axis=-1)
|
||||||
|
|
||||||
|
|
||||||
|
# 3. Generation loop
|
||||||
|
max_new_tokens = 8192
|
||||||
|
generated_tokens = np.array([[]], dtype=np.int64)
|
||||||
|
for i in range(max_new_tokens):
|
||||||
|
inputs_embeds = embed_session.run(None, {'input_ids': input_ids})[0]
|
||||||
|
|
||||||
|
if image_features is None:
|
||||||
|
## Only compute vision features if not already computed
|
||||||
|
image_features = vision_session.run(
|
||||||
|
['image_features'], # List of output names or indices
|
||||||
|
{
|
||||||
|
'pixel_values': inputs['pixel_values'],
|
||||||
|
'pixel_attention_mask': inputs['pixel_attention_mask'].astype(np.bool_)
|
||||||
|
}
|
||||||
|
)[0]
|
||||||
|
|
||||||
|
## Merge text and vision embeddings
|
||||||
|
inputs_embeds[inputs['input_ids'] == image_token_id] = image_features.reshape(-1, image_features.shape[-1])
|
||||||
|
|
||||||
|
logits, *present_key_values = decoder_session.run(None, dict(
|
||||||
|
inputs_embeds=inputs_embeds,
|
||||||
|
attention_mask=attention_mask,
|
||||||
|
position_ids=position_ids,
|
||||||
|
**past_key_values,
|
||||||
|
))
|
||||||
|
|
||||||
|
## Update values for next generation loop
|
||||||
|
input_ids = logits[:, -1].argmax(-1, keepdims=True)
|
||||||
|
attention_mask = np.ones_like(input_ids)
|
||||||
|
position_ids = position_ids[:, -1:] + 1
|
||||||
|
for j, key in enumerate(past_key_values):
|
||||||
|
past_key_values[key] = present_key_values[j]
|
||||||
|
|
||||||
|
generated_tokens = np.concatenate([generated_tokens, input_ids], axis=-1)
|
||||||
|
if (input_ids == eos_token_id).all() or (input_ids == end_of_utterance_id).all():
|
||||||
|
break # Stop predicting
|
||||||
|
|
||||||
|
doctags = processor.batch_decode(
|
||||||
|
generated_tokens,
|
||||||
|
skip_special_tokens=False,
|
||||||
|
)[0].lstrip()
|
||||||
|
|
||||||
|
print(doctags)
|
||||||
|
|
||||||
|
doctags_doc = DocTagsDocument.from_doctags_and_image_pairs([doctags], [image])
|
||||||
|
print(doctags)
|
||||||
|
# create a docling document
|
||||||
|
doc = DoclingDocument.load_from_doctags(doctags_doc, document_name="Document")
|
||||||
|
|
||||||
|
print(doc.export_to_markdown())
|
||||||
|
```
|
||||||
|
</details>
|
||||||
|
|
||||||
|
|
||||||
|
💻 Local inference on Apple Silicon with MLX: [see here](https://huggingface.co/ds4sd/SmolDocling-256M-preview-mlx-bf16)
|
||||||
|
|
||||||
|
## DocTags
|
||||||
|
|
||||||
|
<img src="https://huggingface.co/ds4sd/SmolDocling-256M-preview/resolve/main/assets/doctags_v2.png" width="800" height="auto" alt="Image description">
|
||||||
|
DocTags create a clear and structured system of tags and rules that separate text from the document's structure. This makes things easier for Image-to-Sequence models by reducing confusion. On the other hand, converting directly to formats like HTML or Markdown can be messy—it often loses details, doesn’t clearly show the document’s layout, and increases the number of tokens, making processing less efficient.
|
||||||
|
DocTags are integrated with Docling, which allows export to HTML, Markdown, and JSON. These exports can be offloaded to the CPU, reducing token generation overhead and improving efficiency.
|
||||||
|
|
||||||
|
## Supported Instructions
|
||||||
|
|
||||||
|
<table>
|
||||||
|
<tr>
|
||||||
|
<td><b>Description</b></td>
|
||||||
|
<td><b>Instruction</b></td>
|
||||||
|
<td><b>Comment</b></td>
|
||||||
|
</tr>
|
||||||
|
<tr>
|
||||||
|
<td><b>Full conversion</b></td>
|
||||||
|
<td>Convert this page to docling.</td>
|
||||||
|
<td>DocTags represetation</td>
|
||||||
|
</tr>
|
||||||
|
<tr>
|
||||||
|
<td><b>Chart</b></td>
|
||||||
|
<td>Convert chart to table.</td>
|
||||||
|
<td>(e.g., <chart>)</td>
|
||||||
|
</tr>
|
||||||
|
<tr>
|
||||||
|
<td><b>Formula</b></td>
|
||||||
|
<td>Convert formula to LaTeX.</td>
|
||||||
|
<td>(e.g., <formula>)</td>
|
||||||
|
</tr>
|
||||||
|
<tr>
|
||||||
|
<td><b>Code</b></td>
|
||||||
|
<td>Convert code to text.</td>
|
||||||
|
<td>(e.g., <code>)</td>
|
||||||
|
</tr>
|
||||||
|
<tr>
|
||||||
|
<td><b>Table</b></td>
|
||||||
|
<td>Convert table to OTSL.</td>
|
||||||
|
<td>(e.g., <otsl>) OTSL: <a href="https://arxiv.org/pdf/2305.03393">Lysak et al., 2023</a></td>
|
||||||
|
</tr>
|
||||||
|
<tr>
|
||||||
|
<td rowspan=4><b>Actions and Pipelines</b></td>
|
||||||
|
<td>OCR the text in a specific location: <loc_155><loc_233><loc_206><loc_237></td>
|
||||||
|
<td></td>
|
||||||
|
</tr>
|
||||||
|
<tr>
|
||||||
|
<td>Identify element at: <loc_247><loc_482><10c_252><loc_486></td>
|
||||||
|
<td></td>
|
||||||
|
</tr>
|
||||||
|
<tr>
|
||||||
|
<td>Find all 'text' elements on the page, retrieve all section headers.</td>
|
||||||
|
<td></td>
|
||||||
|
</tr>
|
||||||
|
<tr>
|
||||||
|
<td>Detect footer elements on the page.</td>
|
||||||
|
<td></td>
|
||||||
|
</tr>
|
||||||
|
</table>
|
||||||
|
|
||||||
|
|
||||||
|
#### 📊 Datasets
|
||||||
|
- [SynthCodeNet](https://huggingface.co/datasets/ds4sd/SynthCodeNet)
|
||||||
|
- [SynthFormulaNet](https://huggingface.co/datasets/ds4sd/SynthFormulaNet)
|
||||||
|
- [SynthChartNet](https://huggingface.co/datasets/ds4sd/SynthChartNet)
|
||||||
|
- [DoclingMatix](https://huggingface.co/datasets/HuggingFaceM4/DoclingMatix)
|
||||||
|
|
||||||
|
#### Model Summary
|
||||||
|
|
||||||
|
- **Developed by:** Docling Team, IBM Research
|
||||||
|
- **Model type:** Multi-modal model (image+text)
|
||||||
|
- **Language(s) (NLP):** English
|
||||||
|
- **License:** Apache 2.0
|
||||||
|
- **Architecture:** Based on [Idefics3](https://huggingface.co/HuggingFaceM4/Idefics3-8B-Llama3) (see technical summary)
|
||||||
|
- **Finetuned from model:** Based on [SmolVLM-256M-Instruct](https://huggingface.co/HuggingFaceTB/SmolVLM-256M-Instruct)
|
||||||
|
|
||||||
|
**Repository:** [Docling](https://github.com/docling-project/docling)
|
||||||
|
|
||||||
|
**Paper:** [arXiv](https://arxiv.org/abs/2503.11576)
|
||||||
|
|
||||||
|
**Project Page:** [Hugging Face](https://huggingface.co/ds4sd/SmolDocling-256M-preview)
|
||||||
|
|
||||||
|
**Citation:**
|
||||||
|
```
|
||||||
|
@misc{nassar2025smoldoclingultracompactvisionlanguagemodel,
|
||||||
|
title={SmolDocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion},
|
||||||
|
author={Ahmed Nassar and Andres Marafioti and Matteo Omenetti and Maksym Lysak and Nikolaos Livathinos and Christoph Auer and Lucas Morin and Rafael Teixeira de Lima and Yusik Kim and A. Said Gurbuz and Michele Dolfi and Miquel Farré and Peter W. J. Staar},
|
||||||
|
year={2025},
|
||||||
|
eprint={2503.11576},
|
||||||
|
archivePrefix={arXiv},
|
||||||
|
primaryClass={cs.CV},
|
||||||
|
url={https://arxiv.org/abs/2503.11576},
|
||||||
|
}
|
||||||
|
```
|
||||||
|
**Demo:** [HF Space](https://huggingface.co/spaces/ds4sd/SmolDocling-256M-Demo)
|
||||||
130
added_tokens.json
Normal file
130
added_tokens.json
Normal file
@@ -0,0 +1,130 @@
|
|||||||
|
{
|
||||||
|
"</caption>": 49192,
|
||||||
|
"</chart>": 49248,
|
||||||
|
"</checkbox_selected>": 49211,
|
||||||
|
"</checkbox_unselected>": 49213,
|
||||||
|
"</doctag>": 49230,
|
||||||
|
"</footnote>": 49195,
|
||||||
|
"</form>": 49215,
|
||||||
|
"</formula>": 49197,
|
||||||
|
"</group>": 49228,
|
||||||
|
"</key_": 49243,
|
||||||
|
"</key_value_region>": 49217,
|
||||||
|
"</list_item>": 49199,
|
||||||
|
"</ordered_list>": 49224,
|
||||||
|
"</otsl>": 49209,
|
||||||
|
"</page_footer>": 49201,
|
||||||
|
"</page_header>": 49203,
|
||||||
|
"</paragraph>": 49220,
|
||||||
|
"</picture>": 49205,
|
||||||
|
"</reference>": 49222,
|
||||||
|
"</section_header_level_": 49207,
|
||||||
|
"</smiles>": 49251,
|
||||||
|
"</unordered_list>": 49226,
|
||||||
|
"</value_": 49245,
|
||||||
|
"<caption>": 49191,
|
||||||
|
"<chart>": 49247,
|
||||||
|
"<checkbox_selected>": 49210,
|
||||||
|
"<checkbox_unselected>": 49212,
|
||||||
|
"<ched>": 49239,
|
||||||
|
"<doctag>": 49229,
|
||||||
|
"<ecel>": 49234,
|
||||||
|
"<end_of_utterance>": 49279,
|
||||||
|
"<fake_token_around_image>": 49189,
|
||||||
|
"<fcel>": 49233,
|
||||||
|
"<footnote>": 49193,
|
||||||
|
"<form>": 49214,
|
||||||
|
"<formula>": 49196,
|
||||||
|
"<global-img>": 49152,
|
||||||
|
"<group>": 49227,
|
||||||
|
"<image>": 49190,
|
||||||
|
"<key_": 49242,
|
||||||
|
"<key_value_region>": 49216,
|
||||||
|
"<lcel>": 49235,
|
||||||
|
"<link_": 49246,
|
||||||
|
"<list_item>": 49198,
|
||||||
|
"<loc_": 49218,
|
||||||
|
"<nl>": 49238,
|
||||||
|
"<ordered_list>": 49223,
|
||||||
|
"<otsl>": 49208,
|
||||||
|
"<page_": 49231,
|
||||||
|
"<page_break>": 49249,
|
||||||
|
"<page_footer>": 49200,
|
||||||
|
"<page_header>": 49202,
|
||||||
|
"<paragraph>": 49219,
|
||||||
|
"<picture>": 49204,
|
||||||
|
"<reference>": 49221,
|
||||||
|
"<rhed>": 49240,
|
||||||
|
"<row_1_col_1>": 49153,
|
||||||
|
"<row_1_col_2>": 49154,
|
||||||
|
"<row_1_col_3>": 49155,
|
||||||
|
"<row_1_col_4>": 49156,
|
||||||
|
"<row_1_col_5>": 49157,
|
||||||
|
"<row_1_col_6>": 49158,
|
||||||
|
"<row_2_col_1>": 49159,
|
||||||
|
"<row_2_col_2>": 49160,
|
||||||
|
"<row_2_col_3>": 49161,
|
||||||
|
"<row_2_col_4>": 49162,
|
||||||
|
"<row_2_col_5>": 49163,
|
||||||
|
"<row_2_col_6>": 49164,
|
||||||
|
"<row_3_col_1>": 49165,
|
||||||
|
"<row_3_col_2>": 49166,
|
||||||
|
"<row_3_col_3>": 49167,
|
||||||
|
"<row_3_col_4>": 49168,
|
||||||
|
"<row_3_col_5>": 49169,
|
||||||
|
"<row_3_col_6>": 49170,
|
||||||
|
"<row_4_col_1>": 49171,
|
||||||
|
"<row_4_col_2>": 49172,
|
||||||
|
"<row_4_col_3>": 49173,
|
||||||
|
"<row_4_col_4>": 49174,
|
||||||
|
"<row_4_col_5>": 49175,
|
||||||
|
"<row_4_col_6>": 49176,
|
||||||
|
"<row_5_col_1>": 49177,
|
||||||
|
"<row_5_col_2>": 49178,
|
||||||
|
"<row_5_col_3>": 49179,
|
||||||
|
"<row_5_col_4>": 49180,
|
||||||
|
"<row_5_col_5>": 49181,
|
||||||
|
"<row_5_col_6>": 49182,
|
||||||
|
"<row_6_col_1>": 49183,
|
||||||
|
"<row_6_col_2>": 49184,
|
||||||
|
"<row_6_col_3>": 49185,
|
||||||
|
"<row_6_col_4>": 49186,
|
||||||
|
"<row_6_col_5>": 49187,
|
||||||
|
"<row_6_col_6>": 49188,
|
||||||
|
"<section_header_level_": 49206,
|
||||||
|
"<smiles>": 49250,
|
||||||
|
"<text_break>": 49232,
|
||||||
|
"<ucel>": 49236,
|
||||||
|
"<unordered_list>": 49225,
|
||||||
|
"<value_": 49244,
|
||||||
|
"<xcel>": 49237,
|
||||||
|
"<|reserved_special_token_3|>": 49194,
|
||||||
|
"<|reserved_special_token_50|>": 49241,
|
||||||
|
"<|reserved_special_token_61|>": 49252,
|
||||||
|
"<|reserved_special_token_62|>": 49253,
|
||||||
|
"<|reserved_special_token_63|>": 49254,
|
||||||
|
"<|reserved_special_token_64|>": 49255,
|
||||||
|
"<|reserved_special_token_65|>": 49256,
|
||||||
|
"<|reserved_special_token_66|>": 49257,
|
||||||
|
"<|reserved_special_token_67|>": 49258,
|
||||||
|
"<|reserved_special_token_68|>": 49259,
|
||||||
|
"<|reserved_special_token_69|>": 49260,
|
||||||
|
"<|reserved_special_token_70|>": 49261,
|
||||||
|
"<|reserved_special_token_71|>": 49262,
|
||||||
|
"<|reserved_special_token_72|>": 49263,
|
||||||
|
"<|reserved_special_token_73|>": 49264,
|
||||||
|
"<|reserved_special_token_74|>": 49265,
|
||||||
|
"<|reserved_special_token_75|>": 49266,
|
||||||
|
"<|reserved_special_token_76|>": 49267,
|
||||||
|
"<|reserved_special_token_77|>": 49268,
|
||||||
|
"<|reserved_special_token_78|>": 49269,
|
||||||
|
"<|reserved_special_token_79|>": 49270,
|
||||||
|
"<|reserved_special_token_80|>": 49271,
|
||||||
|
"<|reserved_special_token_81|>": 49272,
|
||||||
|
"<|reserved_special_token_82|>": 49273,
|
||||||
|
"<|reserved_special_token_83|>": 49274,
|
||||||
|
"<|reserved_special_token_84|>": 49275,
|
||||||
|
"<|reserved_special_token_85|>": 49276,
|
||||||
|
"<|reserved_special_token_86|>": 49277,
|
||||||
|
"<|reserved_special_token_87|>": 49278
|
||||||
|
}
|
||||||
BIN
assets/SmolDocling_doctags1.png
Normal file
BIN
assets/SmolDocling_doctags1.png
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 432 KiB |
BIN
assets/doctags_v2.png
Normal file
BIN
assets/doctags_v2.png
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 986 KiB |
3
chat_template.json
Normal file
3
chat_template.json
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
{
|
||||||
|
"chat_template": "<|im_start|>{% for message in messages %}{{message['role'] | capitalize}}{% if message['content'][0]['type'] == 'image' %}{{':'}}{% else %}{{': '}}{% endif %}{% for line in message['content'] %}{% if line['type'] == 'text' %}{{line['text']}}{% elif line['type'] == 'image' %}{{ '<image>' }}{% endif %}{% endfor %}<end_of_utterance>\n{% endfor %}{% if add_generation_prompt %}{{ 'Assistant:' }}{% endif %}"
|
||||||
|
}
|
||||||
143
config.json
Normal file
143
config.json
Normal file
@@ -0,0 +1,143 @@
|
|||||||
|
{
|
||||||
|
"_name_or_path": "/dccstor/ahn_cdip/checkpoints/SmolDocling_250M_DT_ST13/checkpoint-800",
|
||||||
|
"architectures": [
|
||||||
|
"Idefics3ForConditionalGeneration"
|
||||||
|
],
|
||||||
|
"image_token_id": 49190,
|
||||||
|
"model_type": "idefics3",
|
||||||
|
"pad_token_id": 128002,
|
||||||
|
"scale_factor": 4,
|
||||||
|
"text_config": {
|
||||||
|
"_flash_attn_2_enabled": true,
|
||||||
|
"_name_or_path": "None",
|
||||||
|
"architectures": [
|
||||||
|
"VLlama3ForCausalLM"
|
||||||
|
],
|
||||||
|
"head_dim": 64,
|
||||||
|
"hidden_size": 576,
|
||||||
|
"initializer_range": 0.041666666666666664,
|
||||||
|
"intermediate_size": 1536,
|
||||||
|
"is_llama_config": true,
|
||||||
|
"max_position_embeddings": 8192,
|
||||||
|
"model_type": "llama",
|
||||||
|
"neftune_noise_alpha": 0.0,
|
||||||
|
"num_attention_heads": 9,
|
||||||
|
"num_hidden_layers": 30,
|
||||||
|
"num_key_value_heads": 3,
|
||||||
|
"pad_token_id": 2,
|
||||||
|
"perceiver_config": {
|
||||||
|
"_attn_implementation_autoset": false,
|
||||||
|
"_name_or_path": "",
|
||||||
|
"add_cross_attention": false,
|
||||||
|
"architectures": null,
|
||||||
|
"attention_dropout": 0.0,
|
||||||
|
"bad_words_ids": null,
|
||||||
|
"begin_suppress_tokens": null,
|
||||||
|
"bos_token_id": null,
|
||||||
|
"chunk_size_feed_forward": 0,
|
||||||
|
"cross_attention_hidden_size": null,
|
||||||
|
"decoder_start_token_id": null,
|
||||||
|
"diversity_penalty": 0.0,
|
||||||
|
"do_sample": false,
|
||||||
|
"early_stopping": false,
|
||||||
|
"encoder_no_repeat_ngram_size": 0,
|
||||||
|
"eos_token_id": null,
|
||||||
|
"exponential_decay_length_penalty": null,
|
||||||
|
"finetuning_task": null,
|
||||||
|
"forced_bos_token_id": null,
|
||||||
|
"forced_eos_token_id": null,
|
||||||
|
"hidden_act": "silu",
|
||||||
|
"id2label": {
|
||||||
|
"0": "LABEL_0",
|
||||||
|
"1": "LABEL_1"
|
||||||
|
},
|
||||||
|
"is_decoder": false,
|
||||||
|
"is_encoder_decoder": false,
|
||||||
|
"label2id": {
|
||||||
|
"LABEL_0": 0,
|
||||||
|
"LABEL_1": 1
|
||||||
|
},
|
||||||
|
"length_penalty": 1.0,
|
||||||
|
"max_length": 20,
|
||||||
|
"min_length": 0,
|
||||||
|
"model_type": "vllama3",
|
||||||
|
"no_repeat_ngram_size": 0,
|
||||||
|
"num_beam_groups": 1,
|
||||||
|
"num_beams": 1,
|
||||||
|
"num_key_value_heads": 1,
|
||||||
|
"num_return_sequences": 1,
|
||||||
|
"output_attentions": false,
|
||||||
|
"output_hidden_states": false,
|
||||||
|
"output_scores": false,
|
||||||
|
"pad_token_id": null,
|
||||||
|
"prefix": null,
|
||||||
|
"problem_type": null,
|
||||||
|
"pruned_heads": {},
|
||||||
|
"qk_layer_norms_perceiver": false,
|
||||||
|
"remove_invalid_values": false,
|
||||||
|
"repetition_penalty": 1.0,
|
||||||
|
"resampler_depth": 6,
|
||||||
|
"resampler_head_dim": 96,
|
||||||
|
"resampler_n_heads": 16,
|
||||||
|
"resampler_n_latents": 64,
|
||||||
|
"return_dict": true,
|
||||||
|
"return_dict_in_generate": false,
|
||||||
|
"sep_token_id": null,
|
||||||
|
"suppress_tokens": null,
|
||||||
|
"task_specific_params": null,
|
||||||
|
"temperature": 1.0,
|
||||||
|
"tf_legacy_loss": false,
|
||||||
|
"tie_encoder_decoder": false,
|
||||||
|
"tie_word_embeddings": true,
|
||||||
|
"tokenizer_class": null,
|
||||||
|
"top_k": 50,
|
||||||
|
"top_p": 1.0,
|
||||||
|
"torch_dtype": null,
|
||||||
|
"torchscript": false,
|
||||||
|
"transformers_version": "4.46.0",
|
||||||
|
"typical_p": 1.0,
|
||||||
|
"use_bfloat16": false
|
||||||
|
},
|
||||||
|
"pixel_shuffle_factor": 4,
|
||||||
|
"qk_layer_norms": false,
|
||||||
|
"rms_norm_eps": 1e-05,
|
||||||
|
"rope_interleaved": false,
|
||||||
|
"rope_theta": 100000,
|
||||||
|
"torch_dtype": "bfloat16",
|
||||||
|
"transformers.js_config": {
|
||||||
|
"kv_cache_dtype": {
|
||||||
|
"fp16": "float16",
|
||||||
|
"q4f16": "float16"
|
||||||
|
}
|
||||||
|
},
|
||||||
|
"use_resampler": false,
|
||||||
|
"vocab_size": 49280
|
||||||
|
},
|
||||||
|
"tie_word_embeddings": false,
|
||||||
|
"torch_dtype": "bfloat16",
|
||||||
|
"transformers_version": "4.50.0.dev0",
|
||||||
|
"transformers.js_config": {
|
||||||
|
"kv_cache_dtype": {
|
||||||
|
"q4f16": "float16",
|
||||||
|
"fp16": "float16"
|
||||||
|
}
|
||||||
|
},
|
||||||
|
"use_cache": true,
|
||||||
|
"vision_config": {
|
||||||
|
"hidden_size": 768,
|
||||||
|
"image_size": 512,
|
||||||
|
"max_image_size": {
|
||||||
|
"longest_edge": 512
|
||||||
|
},
|
||||||
|
"model_type": "idefics3_vision",
|
||||||
|
"num_attention_heads": 12,
|
||||||
|
"patch_size": 16,
|
||||||
|
"size": {
|
||||||
|
"longest_edge": 2048
|
||||||
|
},
|
||||||
|
"tie_word_embeddings": false,
|
||||||
|
"torch_dtype": "bfloat16",
|
||||||
|
"use_base_siglip": true
|
||||||
|
},
|
||||||
|
"vocab_size": 49280
|
||||||
|
}
|
||||||
1
configuration.json
Normal file
1
configuration.json
Normal file
@@ -0,0 +1 @@
|
|||||||
|
{"framework": "pytorch", "task": "image-text-to-text", "allow_remote": true}
|
||||||
7
generation_config.json
Normal file
7
generation_config.json
Normal file
@@ -0,0 +1,7 @@
|
|||||||
|
{
|
||||||
|
"_from_model_config": true,
|
||||||
|
"bos_token_id": 0,
|
||||||
|
"eos_token_id": 49279,
|
||||||
|
"pad_token_id": 2,
|
||||||
|
"transformers_version": "4.50.0.dev0"
|
||||||
|
}
|
||||||
48901
merges.txt
Normal file
48901
merges.txt
Normal file
File diff suppressed because it is too large
Load Diff
3
model.safetensors
Normal file
3
model.safetensors
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:cdcdf5d823c5684029c7d8e52177cf10f9034b3aba6577549cfb1a9ce36ad0a2
|
||||||
|
size 513028808
|
||||||
3
onnx/decoder_model_merged.onnx
Normal file
3
onnx/decoder_model_merged.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:c75155e5e121b9d89bf5435d396ccd9e817572d8a0e4589e1c0eb5574bb3c69e
|
||||||
|
size 540972236
|
||||||
3
onnx/decoder_model_merged_bnb4.onnx
Normal file
3
onnx/decoder_model_merged_bnb4.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:51ef8fa60e201a412c1a0b24333bfa2a6c73d4f45e58b9e2411866cfc6fc707b
|
||||||
|
size 78486831
|
||||||
3
onnx/decoder_model_merged_fp16.onnx
Normal file
3
onnx/decoder_model_merged_fp16.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:ccd0761dc1e061c706381fb345a01977d4ec4db2ee97d3acb6886f65fbe2ba81
|
||||||
|
size 270764878
|
||||||
3
onnx/decoder_model_merged_int8.onnx
Normal file
3
onnx/decoder_model_merged_int8.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:9fdce1ec841ad5b730d2c220ff2147f2114cecf7b5518c27cc050c29bf75e997
|
||||||
|
size 137553254
|
||||||
3
onnx/decoder_model_merged_q4.onnx
Normal file
3
onnx/decoder_model_merged_q4.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:ef6045cf8e613277a745f5b56eaf1cc2dda00c64d984acaa005fa5ef706efcc5
|
||||||
|
size 86894835
|
||||||
3
onnx/decoder_model_merged_q4f16.onnx
Normal file
3
onnx/decoder_model_merged_q4f16.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:e9cda6976be0f2b446a01ace90aa04a520116b4641e544cbc74dd8a70335b606
|
||||||
|
size 77385255
|
||||||
3
onnx/decoder_model_merged_quantized.onnx
Normal file
3
onnx/decoder_model_merged_quantized.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:88d2ef3239d292224065443e7dc6a8b87195db94a38280e23bd4b5ca18c1d3b5
|
||||||
|
size 137553365
|
||||||
3
onnx/decoder_model_merged_uint8.onnx
Normal file
3
onnx/decoder_model_merged_uint8.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:88d2ef3239d292224065443e7dc6a8b87195db94a38280e23bd4b5ca18c1d3b5
|
||||||
|
size 137553365
|
||||||
3
onnx/embed_tokens.onnx
Normal file
3
onnx/embed_tokens.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:816939a02a48a540330655f6cc1a33cb47b2021aec0ebb356a016c50b826f3f6
|
||||||
|
size 113541419
|
||||||
3
onnx/embed_tokens_bnb4.onnx
Normal file
3
onnx/embed_tokens_bnb4.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:7b6ad5d0ac29dc4bd893b61606f760fac40481a4eecbcb318502b05e3562aa37
|
||||||
|
size 113541438
|
||||||
3
onnx/embed_tokens_fp16.onnx
Normal file
3
onnx/embed_tokens_fp16.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:f08a824993fe5ea9d5affd34a986a3700c4c5cb0fd21993182c03a58c4a0ee3a
|
||||||
|
size 56770887
|
||||||
3
onnx/embed_tokens_int8.onnx
Normal file
3
onnx/embed_tokens_int8.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:74e69c670cb770eb3b430c7fdd770a9e4e0f174edef568e6aa67df4276c398e5
|
||||||
|
size 28385824
|
||||||
3
onnx/embed_tokens_q4.onnx
Normal file
3
onnx/embed_tokens_q4.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:7b6ad5d0ac29dc4bd893b61606f760fac40481a4eecbcb318502b05e3562aa37
|
||||||
|
size 113541438
|
||||||
3
onnx/embed_tokens_q4f16.onnx
Normal file
3
onnx/embed_tokens_q4f16.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:664289fb111a87112cf97c9ac4d9122119740bc5da3a6eb8305fa1063dec65a4
|
||||||
|
size 56770906
|
||||||
3
onnx/embed_tokens_quantized.onnx
Normal file
3
onnx/embed_tokens_quantized.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:74e69c670cb770eb3b430c7fdd770a9e4e0f174edef568e6aa67df4276c398e5
|
||||||
|
size 28385824
|
||||||
3
onnx/embed_tokens_uint8.onnx
Normal file
3
onnx/embed_tokens_uint8.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:74e69c670cb770eb3b430c7fdd770a9e4e0f174edef568e6aa67df4276c398e5
|
||||||
|
size 28385824
|
||||||
3
onnx/vision_encoder.onnx
Normal file
3
onnx/vision_encoder.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:9c97aafac0543632c778938e2870afb73a81c7cc3a3e01e84b43a096bd668820
|
||||||
|
size 374316454
|
||||||
3
onnx/vision_encoder_bnb4.onnx
Normal file
3
onnx/vision_encoder_bnb4.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:fcc3b8bf6e0b12be05c5ce66adc5d504fb0efb4a00dbfc53a8f8dbc1cca1516e
|
||||||
|
size 58034696
|
||||||
3
onnx/vision_encoder_fp16.onnx
Normal file
3
onnx/vision_encoder_fp16.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:f86f88e185e231b91232147206771e267d8ab038aa4e30fb29e376d9ec4db4fb
|
||||||
|
size 187295327
|
||||||
3
onnx/vision_encoder_int8.onnx
Normal file
3
onnx/vision_encoder_int8.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:3251860101c63f90c44ec475c874346c461de48421aacc13b5f5741004282240
|
||||||
|
size 94247884
|
||||||
3
onnx/vision_encoder_q4.onnx
Normal file
3
onnx/vision_encoder_q4.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:fbcfa73fba2477b564a364123bc066922c9e692a9389c24eb904009f922e9aa0
|
||||||
|
size 63784944
|
||||||
3
onnx/vision_encoder_q4f16.onnx
Normal file
3
onnx/vision_encoder_q4f16.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:c0e3c83edfaf3c946e592f50d9b238eb422c0c4ef62643883a46d7c5ebb7284a
|
||||||
|
size 55038218
|
||||||
3
onnx/vision_encoder_quantized.onnx
Normal file
3
onnx/vision_encoder_quantized.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:b5e42cf6c8719dba60582c3722cba72673de5e2b739c01e5147bc34d9193ae20
|
||||||
|
size 94247926
|
||||||
3
onnx/vision_encoder_uint8.onnx
Normal file
3
onnx/vision_encoder_uint8.onnx
Normal file
@@ -0,0 +1,3 @@
|
|||||||
|
version https://git-lfs.github.com/spec/v1
|
||||||
|
oid sha256:b5e42cf6c8719dba60582c3722cba72673de5e2b739c01e5147bc34d9193ae20
|
||||||
|
size 94247926
|
||||||
28
preprocessor_config.json
Normal file
28
preprocessor_config.json
Normal file
@@ -0,0 +1,28 @@
|
|||||||
|
{
|
||||||
|
"do_convert_rgb": true,
|
||||||
|
"do_image_splitting": true,
|
||||||
|
"do_normalize": true,
|
||||||
|
"do_pad": true,
|
||||||
|
"do_rescale": true,
|
||||||
|
"do_resize": true,
|
||||||
|
"image_mean": [
|
||||||
|
0.5,
|
||||||
|
0.5,
|
||||||
|
0.5
|
||||||
|
],
|
||||||
|
"image_processor_type": "Idefics3ImageProcessor",
|
||||||
|
"image_std": [
|
||||||
|
0.5,
|
||||||
|
0.5,
|
||||||
|
0.5
|
||||||
|
],
|
||||||
|
"max_image_size": {
|
||||||
|
"longest_edge": 512
|
||||||
|
},
|
||||||
|
"processor_class": "Idefics3Processor",
|
||||||
|
"resample": 1,
|
||||||
|
"rescale_factor": 0.00392156862745098,
|
||||||
|
"size": {
|
||||||
|
"longest_edge": 2048
|
||||||
|
}
|
||||||
|
}
|
||||||
4
processor_config.json
Normal file
4
processor_config.json
Normal file
@@ -0,0 +1,4 @@
|
|||||||
|
{
|
||||||
|
"image_seq_len": 64,
|
||||||
|
"processor_class": "Idefics3Processor"
|
||||||
|
}
|
||||||
53
special_tokens_map.json
Normal file
53
special_tokens_map.json
Normal file
@@ -0,0 +1,53 @@
|
|||||||
|
{
|
||||||
|
"additional_special_tokens": [
|
||||||
|
{
|
||||||
|
"content": "<fake_token_around_image>",
|
||||||
|
"lstrip": false,
|
||||||
|
"normalized": false,
|
||||||
|
"rstrip": false,
|
||||||
|
"single_word": false
|
||||||
|
},
|
||||||
|
{
|
||||||
|
"content": "<image>",
|
||||||
|
"lstrip": false,
|
||||||
|
"normalized": false,
|
||||||
|
"rstrip": false,
|
||||||
|
"single_word": false
|
||||||
|
},
|
||||||
|
{
|
||||||
|
"content": "<end_of_utterance>",
|
||||||
|
"lstrip": false,
|
||||||
|
"normalized": false,
|
||||||
|
"rstrip": false,
|
||||||
|
"single_word": false
|
||||||
|
}
|
||||||
|
],
|
||||||
|
"bos_token": {
|
||||||
|
"content": "<|im_start|>",
|
||||||
|
"lstrip": false,
|
||||||
|
"normalized": false,
|
||||||
|
"rstrip": false,
|
||||||
|
"single_word": false
|
||||||
|
},
|
||||||
|
"eos_token": {
|
||||||
|
"content": "<|im_end|>",
|
||||||
|
"lstrip": false,
|
||||||
|
"normalized": false,
|
||||||
|
"rstrip": false,
|
||||||
|
"single_word": false
|
||||||
|
},
|
||||||
|
"pad_token": {
|
||||||
|
"content": "<|im_end|>",
|
||||||
|
"lstrip": false,
|
||||||
|
"normalized": false,
|
||||||
|
"rstrip": false,
|
||||||
|
"single_word": false
|
||||||
|
},
|
||||||
|
"unk_token": {
|
||||||
|
"content": "<|endoftext|>",
|
||||||
|
"lstrip": false,
|
||||||
|
"normalized": false,
|
||||||
|
"rstrip": false,
|
||||||
|
"single_word": false
|
||||||
|
}
|
||||||
|
}
|
||||||
246113
tokenizer.json
Normal file
246113
tokenizer.json
Normal file
File diff suppressed because it is too large
Load Diff
1189
tokenizer_config.json
Normal file
1189
tokenizer_config.json
Normal file
File diff suppressed because it is too large
Load Diff
1
vocab.json
Normal file
1
vocab.json
Normal file
File diff suppressed because one or more lines are too long
760
zero_to_fp32.py
Normal file
760
zero_to_fp32.py
Normal file
@@ -0,0 +1,760 @@
|
|||||||
|
#!/usr/bin/env python
|
||||||
|
|
||||||
|
# Copyright (c) Microsoft Corporation.
|
||||||
|
# SPDX-License-Identifier: Apache-2.0
|
||||||
|
|
||||||
|
# DeepSpeed Team
|
||||||
|
|
||||||
|
# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
|
||||||
|
# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
|
||||||
|
# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
|
||||||
|
# application.
|
||||||
|
#
|
||||||
|
# example:
|
||||||
|
# python zero_to_fp32.py . output_dir/
|
||||||
|
# or
|
||||||
|
# python zero_to_fp32.py . output_dir/ --safe_serialization
|
||||||
|
|
||||||
|
import argparse
|
||||||
|
import torch
|
||||||
|
import glob
|
||||||
|
import math
|
||||||
|
import os
|
||||||
|
import re
|
||||||
|
import gc
|
||||||
|
import json
|
||||||
|
import numpy as np
|
||||||
|
from tqdm import tqdm
|
||||||
|
from collections import OrderedDict
|
||||||
|
from dataclasses import dataclass
|
||||||
|
|
||||||
|
# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
|
||||||
|
# DeepSpeed data structures it has to be available in the current python environment.
|
||||||
|
from deepspeed.utils import logger
|
||||||
|
from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
|
||||||
|
FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
|
||||||
|
FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
|
||||||
|
|
||||||
|
|
||||||
|
@dataclass
|
||||||
|
class zero_model_state:
|
||||||
|
buffers: dict()
|
||||||
|
param_shapes: dict()
|
||||||
|
shared_params: list
|
||||||
|
ds_version: int
|
||||||
|
frozen_param_shapes: dict()
|
||||||
|
frozen_param_fragments: dict()
|
||||||
|
|
||||||
|
|
||||||
|
debug = 0
|
||||||
|
|
||||||
|
# load to cpu
|
||||||
|
device = torch.device('cpu')
|
||||||
|
|
||||||
|
|
||||||
|
def atoi(text):
|
||||||
|
return int(text) if text.isdigit() else text
|
||||||
|
|
||||||
|
|
||||||
|
def natural_keys(text):
|
||||||
|
'''
|
||||||
|
alist.sort(key=natural_keys) sorts in human order
|
||||||
|
http://nedbatchelder.com/blog/200712/human_sorting.html
|
||||||
|
(See Toothy's implementation in the comments)
|
||||||
|
'''
|
||||||
|
return [atoi(c) for c in re.split(r'(\d+)', text)]
|
||||||
|
|
||||||
|
|
||||||
|
def get_model_state_file(checkpoint_dir, zero_stage):
|
||||||
|
if not os.path.isdir(checkpoint_dir):
|
||||||
|
raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
|
||||||
|
|
||||||
|
# there should be only one file
|
||||||
|
if zero_stage <= 2:
|
||||||
|
file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
|
||||||
|
elif zero_stage == 3:
|
||||||
|
file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
|
||||||
|
|
||||||
|
if not os.path.exists(file):
|
||||||
|
raise FileNotFoundError(f"can't find model states file at '{file}'")
|
||||||
|
|
||||||
|
return file
|
||||||
|
|
||||||
|
|
||||||
|
def get_checkpoint_files(checkpoint_dir, glob_pattern):
|
||||||
|
# XXX: need to test that this simple glob rule works for multi-node setup too
|
||||||
|
ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
|
||||||
|
|
||||||
|
if len(ckpt_files) == 0:
|
||||||
|
raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
|
||||||
|
|
||||||
|
return ckpt_files
|
||||||
|
|
||||||
|
|
||||||
|
def get_optim_files(checkpoint_dir):
|
||||||
|
return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
|
||||||
|
|
||||||
|
|
||||||
|
def get_model_state_files(checkpoint_dir):
|
||||||
|
return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
|
||||||
|
|
||||||
|
|
||||||
|
def parse_model_states(files):
|
||||||
|
zero_model_states = []
|
||||||
|
for file in files:
|
||||||
|
state_dict = torch.load(file, map_location=device, weights_only=False)
|
||||||
|
|
||||||
|
if BUFFER_NAMES not in state_dict:
|
||||||
|
raise ValueError(f"{file} is not a model state checkpoint")
|
||||||
|
buffer_names = state_dict[BUFFER_NAMES]
|
||||||
|
if debug:
|
||||||
|
print("Found buffers:", buffer_names)
|
||||||
|
|
||||||
|
# recover just the buffers while restoring them to fp32 if they were saved in fp16
|
||||||
|
buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
|
||||||
|
param_shapes = state_dict[PARAM_SHAPES]
|
||||||
|
|
||||||
|
# collect parameters that are included in param_shapes
|
||||||
|
param_names = []
|
||||||
|
for s in param_shapes:
|
||||||
|
for name in s.keys():
|
||||||
|
param_names.append(name)
|
||||||
|
|
||||||
|
# update with frozen parameters
|
||||||
|
frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
|
||||||
|
if frozen_param_shapes is not None:
|
||||||
|
if debug:
|
||||||
|
print(f"Found frozen_param_shapes: {frozen_param_shapes}")
|
||||||
|
param_names += list(frozen_param_shapes.keys())
|
||||||
|
|
||||||
|
# handle shared params
|
||||||
|
shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
|
||||||
|
|
||||||
|
ds_version = state_dict.get(DS_VERSION, None)
|
||||||
|
|
||||||
|
frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
|
||||||
|
|
||||||
|
z_model_state = zero_model_state(buffers=buffers,
|
||||||
|
param_shapes=param_shapes,
|
||||||
|
shared_params=shared_params,
|
||||||
|
ds_version=ds_version,
|
||||||
|
frozen_param_shapes=frozen_param_shapes,
|
||||||
|
frozen_param_fragments=frozen_param_fragments)
|
||||||
|
zero_model_states.append(z_model_state)
|
||||||
|
|
||||||
|
return zero_model_states
|
||||||
|
|
||||||
|
|
||||||
|
def parse_optim_states(files, ds_checkpoint_dir):
|
||||||
|
total_files = len(files)
|
||||||
|
state_dicts = []
|
||||||
|
for f in tqdm(files, desc='Loading checkpoint shards'):
|
||||||
|
state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
|
||||||
|
# immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
|
||||||
|
# and also handle the case where it was already removed by another helper script
|
||||||
|
state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
|
||||||
|
state_dicts.append(state_dict)
|
||||||
|
|
||||||
|
if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
|
||||||
|
raise ValueError(f"{files[0]} is not a zero checkpoint")
|
||||||
|
zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
|
||||||
|
world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
|
||||||
|
|
||||||
|
# For ZeRO-2 each param group can have different partition_count as data parallelism for expert
|
||||||
|
# parameters can be different from data parallelism for non-expert parameters. So we can just
|
||||||
|
# use the max of the partition_count to get the dp world_size.
|
||||||
|
|
||||||
|
if type(world_size) is list:
|
||||||
|
world_size = max(world_size)
|
||||||
|
|
||||||
|
if world_size != total_files:
|
||||||
|
raise ValueError(
|
||||||
|
f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
|
||||||
|
"Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
|
||||||
|
)
|
||||||
|
|
||||||
|
# the groups are named differently in each stage
|
||||||
|
if zero_stage <= 2:
|
||||||
|
fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
|
||||||
|
elif zero_stage == 3:
|
||||||
|
fp32_groups_key = FP32_FLAT_GROUPS
|
||||||
|
else:
|
||||||
|
raise ValueError(f"unknown zero stage {zero_stage}")
|
||||||
|
|
||||||
|
fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
|
||||||
|
return zero_stage, world_size, fp32_flat_groups
|
||||||
|
|
||||||
|
|
||||||
|
def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
|
||||||
|
"""
|
||||||
|
Returns fp32 state_dict reconstructed from ds checkpoint
|
||||||
|
|
||||||
|
Args:
|
||||||
|
- ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
|
||||||
|
|
||||||
|
"""
|
||||||
|
print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
|
||||||
|
|
||||||
|
optim_files = get_optim_files(ds_checkpoint_dir)
|
||||||
|
zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
|
||||||
|
print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
|
||||||
|
|
||||||
|
model_files = get_model_state_files(ds_checkpoint_dir)
|
||||||
|
|
||||||
|
zero_model_states = parse_model_states(model_files)
|
||||||
|
print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
|
||||||
|
|
||||||
|
if zero_stage <= 2:
|
||||||
|
return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
|
||||||
|
exclude_frozen_parameters)
|
||||||
|
elif zero_stage == 3:
|
||||||
|
return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
|
||||||
|
exclude_frozen_parameters)
|
||||||
|
|
||||||
|
|
||||||
|
def _zero2_merge_frozen_params(state_dict, zero_model_states):
|
||||||
|
if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
|
||||||
|
return
|
||||||
|
|
||||||
|
frozen_param_shapes = zero_model_states[0].frozen_param_shapes
|
||||||
|
frozen_param_fragments = zero_model_states[0].frozen_param_fragments
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
num_elem = sum(s.numel() for s in frozen_param_shapes.values())
|
||||||
|
print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
|
||||||
|
|
||||||
|
wanted_params = len(frozen_param_shapes)
|
||||||
|
wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
|
||||||
|
avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
|
||||||
|
print(f'Frozen params: Have {avail_numel} numels to process.')
|
||||||
|
print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
|
||||||
|
|
||||||
|
total_params = 0
|
||||||
|
total_numel = 0
|
||||||
|
for name, shape in frozen_param_shapes.items():
|
||||||
|
total_params += 1
|
||||||
|
unpartitioned_numel = shape.numel()
|
||||||
|
total_numel += unpartitioned_numel
|
||||||
|
|
||||||
|
state_dict[name] = frozen_param_fragments[name]
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
|
||||||
|
|
||||||
|
print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
|
||||||
|
|
||||||
|
|
||||||
|
def _has_callable(obj, fn):
|
||||||
|
attr = getattr(obj, fn, None)
|
||||||
|
return callable(attr)
|
||||||
|
|
||||||
|
|
||||||
|
def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
|
||||||
|
param_shapes = zero_model_states[0].param_shapes
|
||||||
|
|
||||||
|
# Reconstruction protocol:
|
||||||
|
#
|
||||||
|
# XXX: document this
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
for i in range(world_size):
|
||||||
|
for j in range(len(fp32_flat_groups[0])):
|
||||||
|
print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
|
||||||
|
|
||||||
|
# XXX: memory usage doubles here (zero2)
|
||||||
|
num_param_groups = len(fp32_flat_groups[0])
|
||||||
|
merged_single_partition_of_fp32_groups = []
|
||||||
|
for i in range(num_param_groups):
|
||||||
|
merged_partitions = [sd[i] for sd in fp32_flat_groups]
|
||||||
|
full_single_fp32_vector = torch.cat(merged_partitions, 0)
|
||||||
|
merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
|
||||||
|
avail_numel = sum(
|
||||||
|
[full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
wanted_params = sum([len(shapes) for shapes in param_shapes])
|
||||||
|
wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
|
||||||
|
# not asserting if there is a mismatch due to possible padding
|
||||||
|
print(f"Have {avail_numel} numels to process.")
|
||||||
|
print(f"Need {wanted_numel} numels in {wanted_params} params.")
|
||||||
|
|
||||||
|
# params
|
||||||
|
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
|
||||||
|
# out-of-core computing solution
|
||||||
|
total_numel = 0
|
||||||
|
total_params = 0
|
||||||
|
for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
|
||||||
|
offset = 0
|
||||||
|
avail_numel = full_single_fp32_vector.numel()
|
||||||
|
for name, shape in shapes.items():
|
||||||
|
|
||||||
|
unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
|
||||||
|
total_numel += unpartitioned_numel
|
||||||
|
total_params += 1
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
|
||||||
|
state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
|
||||||
|
offset += unpartitioned_numel
|
||||||
|
|
||||||
|
# Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
|
||||||
|
# avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
|
||||||
|
# paddings performed in the code it's almost impossible to predict the exact numbers w/o the
|
||||||
|
# live optimizer object, so we are checking that the numbers are within the right range
|
||||||
|
align_to = 2 * world_size
|
||||||
|
|
||||||
|
def zero2_align(x):
|
||||||
|
return align_to * math.ceil(x / align_to)
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
print(f"original offset={offset}, avail_numel={avail_numel}")
|
||||||
|
|
||||||
|
offset = zero2_align(offset)
|
||||||
|
avail_numel = zero2_align(avail_numel)
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
print(f"aligned offset={offset}, avail_numel={avail_numel}")
|
||||||
|
|
||||||
|
# Sanity check
|
||||||
|
if offset != avail_numel:
|
||||||
|
raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
|
||||||
|
|
||||||
|
print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
|
||||||
|
|
||||||
|
|
||||||
|
def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
|
||||||
|
exclude_frozen_parameters):
|
||||||
|
state_dict = OrderedDict()
|
||||||
|
|
||||||
|
# buffers
|
||||||
|
buffers = zero_model_states[0].buffers
|
||||||
|
state_dict.update(buffers)
|
||||||
|
if debug:
|
||||||
|
print(f"added {len(buffers)} buffers")
|
||||||
|
|
||||||
|
if not exclude_frozen_parameters:
|
||||||
|
_zero2_merge_frozen_params(state_dict, zero_model_states)
|
||||||
|
|
||||||
|
_zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
|
||||||
|
|
||||||
|
# recover shared parameters
|
||||||
|
for pair in zero_model_states[0].shared_params:
|
||||||
|
if pair[1] in state_dict:
|
||||||
|
state_dict[pair[0]] = state_dict[pair[1]]
|
||||||
|
|
||||||
|
return state_dict
|
||||||
|
|
||||||
|
|
||||||
|
def zero3_partitioned_param_info(unpartitioned_numel, world_size):
|
||||||
|
remainder = unpartitioned_numel % world_size
|
||||||
|
padding_numel = (world_size - remainder) if remainder else 0
|
||||||
|
partitioned_numel = math.ceil(unpartitioned_numel / world_size)
|
||||||
|
return partitioned_numel, padding_numel
|
||||||
|
|
||||||
|
|
||||||
|
def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
|
||||||
|
if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
|
||||||
|
return
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
for i in range(world_size):
|
||||||
|
num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
|
||||||
|
print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
|
||||||
|
|
||||||
|
frozen_param_shapes = zero_model_states[0].frozen_param_shapes
|
||||||
|
wanted_params = len(frozen_param_shapes)
|
||||||
|
wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
|
||||||
|
avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
|
||||||
|
print(f'Frozen params: Have {avail_numel} numels to process.')
|
||||||
|
print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
|
||||||
|
|
||||||
|
total_params = 0
|
||||||
|
total_numel = 0
|
||||||
|
for name, shape in zero_model_states[0].frozen_param_shapes.items():
|
||||||
|
total_params += 1
|
||||||
|
unpartitioned_numel = shape.numel()
|
||||||
|
total_numel += unpartitioned_numel
|
||||||
|
|
||||||
|
param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
|
||||||
|
state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
|
||||||
|
|
||||||
|
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
print(
|
||||||
|
f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
|
||||||
|
)
|
||||||
|
|
||||||
|
print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
|
||||||
|
|
||||||
|
|
||||||
|
class GatheredTensor:
|
||||||
|
"""
|
||||||
|
A pseudo tensor that collects partitioned weights.
|
||||||
|
It is more memory efficient when there are multiple groups.
|
||||||
|
"""
|
||||||
|
|
||||||
|
def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
|
||||||
|
self.flat_groups = flat_groups
|
||||||
|
self.flat_groups_offset = flat_groups_offset
|
||||||
|
self.offset = offset
|
||||||
|
self.partitioned_numel = partitioned_numel
|
||||||
|
self.shape = shape
|
||||||
|
self.dtype = self.flat_groups[0][0].dtype
|
||||||
|
|
||||||
|
def contiguous(self):
|
||||||
|
"""
|
||||||
|
Merge partitioned weights from flat_groups into a single tensor.
|
||||||
|
"""
|
||||||
|
end_idx = self.offset + self.partitioned_numel
|
||||||
|
world_size = len(self.flat_groups)
|
||||||
|
pad_flat_param_chunks = []
|
||||||
|
|
||||||
|
for rank_i in range(world_size):
|
||||||
|
# for each rank, we need to collect weights from related group/groups
|
||||||
|
flat_groups_at_rank_i = self.flat_groups[rank_i]
|
||||||
|
start_group_id = None
|
||||||
|
end_group_id = None
|
||||||
|
for group_id in range(len(self.flat_groups_offset)):
|
||||||
|
if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
|
||||||
|
start_group_id = group_id
|
||||||
|
if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
|
||||||
|
end_group_id = group_id
|
||||||
|
break
|
||||||
|
# collect weights from related group/groups
|
||||||
|
for group_id in range(start_group_id, end_group_id + 1):
|
||||||
|
flat_tensor = flat_groups_at_rank_i[group_id]
|
||||||
|
start_offset = self.offset - self.flat_groups_offset[group_id]
|
||||||
|
end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
|
||||||
|
pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
|
||||||
|
|
||||||
|
# collect weights from all ranks
|
||||||
|
pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
|
||||||
|
param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
|
||||||
|
return param
|
||||||
|
|
||||||
|
|
||||||
|
def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
|
||||||
|
param_shapes = zero_model_states[0].param_shapes
|
||||||
|
avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
|
||||||
|
|
||||||
|
# Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
|
||||||
|
# param, re-consolidating each param, while dealing with padding if any
|
||||||
|
|
||||||
|
# merge list of dicts, preserving order
|
||||||
|
param_shapes = {k: v for d in param_shapes for k, v in d.items()}
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
for i in range(world_size):
|
||||||
|
print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
|
||||||
|
|
||||||
|
wanted_params = len(param_shapes)
|
||||||
|
wanted_numel = sum(shape.numel() for shape in param_shapes.values())
|
||||||
|
# not asserting if there is a mismatch due to possible padding
|
||||||
|
avail_numel = fp32_flat_groups[0].numel() * world_size
|
||||||
|
print(f"Trainable params: Have {avail_numel} numels to process.")
|
||||||
|
print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
|
||||||
|
|
||||||
|
# params
|
||||||
|
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
|
||||||
|
# out-of-core computing solution
|
||||||
|
offset = 0
|
||||||
|
total_numel = 0
|
||||||
|
total_params = 0
|
||||||
|
flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
|
||||||
|
for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
|
||||||
|
unpartitioned_numel = shape.numel()
|
||||||
|
total_numel += unpartitioned_numel
|
||||||
|
total_params += 1
|
||||||
|
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
|
||||||
|
|
||||||
|
if debug:
|
||||||
|
print(
|
||||||
|
f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
|
||||||
|
)
|
||||||
|
|
||||||
|
# memory efficient tensor
|
||||||
|
tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
|
||||||
|
state_dict[name] = tensor
|
||||||
|
offset += partitioned_numel
|
||||||
|
|
||||||
|
offset *= world_size
|
||||||
|
|
||||||
|
# Sanity check
|
||||||
|
if offset != avail_numel:
|
||||||
|
raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
|
||||||
|
|
||||||
|
print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
|
||||||
|
|
||||||
|
|
||||||
|
def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
|
||||||
|
exclude_frozen_parameters):
|
||||||
|
state_dict = OrderedDict()
|
||||||
|
|
||||||
|
# buffers
|
||||||
|
buffers = zero_model_states[0].buffers
|
||||||
|
state_dict.update(buffers)
|
||||||
|
if debug:
|
||||||
|
print(f"added {len(buffers)} buffers")
|
||||||
|
|
||||||
|
if not exclude_frozen_parameters:
|
||||||
|
_zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
|
||||||
|
|
||||||
|
_zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
|
||||||
|
|
||||||
|
# recover shared parameters
|
||||||
|
for pair in zero_model_states[0].shared_params:
|
||||||
|
if pair[1] in state_dict:
|
||||||
|
state_dict[pair[0]] = state_dict[pair[1]]
|
||||||
|
|
||||||
|
return state_dict
|
||||||
|
|
||||||
|
|
||||||
|
def to_torch_tensor(state_dict, return_empty_tensor=False):
|
||||||
|
"""
|
||||||
|
Convert state_dict of GatheredTensor to torch tensor
|
||||||
|
"""
|
||||||
|
torch_state_dict = {}
|
||||||
|
converted_tensors = {}
|
||||||
|
for name, tensor in state_dict.items():
|
||||||
|
tensor_id = id(tensor)
|
||||||
|
if tensor_id in converted_tensors: # shared tensors
|
||||||
|
shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
|
||||||
|
torch_state_dict[name] = shared_tensor
|
||||||
|
else:
|
||||||
|
converted_tensors[tensor_id] = name
|
||||||
|
if return_empty_tensor:
|
||||||
|
torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
|
||||||
|
else:
|
||||||
|
torch_state_dict[name] = tensor.contiguous()
|
||||||
|
return torch_state_dict
|
||||||
|
|
||||||
|
|
||||||
|
def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
|
||||||
|
tag=None,
|
||||||
|
exclude_frozen_parameters=False,
|
||||||
|
lazy_mode=False):
|
||||||
|
"""
|
||||||
|
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
|
||||||
|
``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
|
||||||
|
via a model hub.
|
||||||
|
|
||||||
|
Args:
|
||||||
|
- ``checkpoint_dir``: path to the desired checkpoint folder
|
||||||
|
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
|
||||||
|
- ``exclude_frozen_parameters``: exclude frozen parameters
|
||||||
|
- ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
|
||||||
|
Convert the pesduo tensor to torch tensor by ``.contiguous()``
|
||||||
|
|
||||||
|
Returns:
|
||||||
|
- pytorch ``state_dict``
|
||||||
|
|
||||||
|
A typical usage might be ::
|
||||||
|
|
||||||
|
from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
|
||||||
|
# do the training and checkpoint saving
|
||||||
|
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
|
||||||
|
model = model.cpu() # move to cpu
|
||||||
|
model.load_state_dict(state_dict)
|
||||||
|
# submit to model hub or save the model to share with others
|
||||||
|
|
||||||
|
In this example the ``model`` will no longer be usable in the deepspeed context of the same
|
||||||
|
application. i.e. you will need to re-initialize the deepspeed engine, since
|
||||||
|
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
|
||||||
|
|
||||||
|
If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
|
||||||
|
|
||||||
|
Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
|
||||||
|
You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
|
||||||
|
the checkpoint. Or you can load state_dict in lazy mode ::
|
||||||
|
|
||||||
|
from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
|
||||||
|
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
|
||||||
|
for name, lazy_tensor in state_dict.item():
|
||||||
|
tensor = lazy_tensor.contiguous() # to cpu
|
||||||
|
print(name, tensor)
|
||||||
|
# del tensor to release memory if it no longer in use
|
||||||
|
"""
|
||||||
|
if tag is None:
|
||||||
|
latest_path = os.path.join(checkpoint_dir, 'latest')
|
||||||
|
if os.path.isfile(latest_path):
|
||||||
|
with open(latest_path, 'r') as fd:
|
||||||
|
tag = fd.read().strip()
|
||||||
|
else:
|
||||||
|
raise ValueError(f"Unable to find 'latest' file at {latest_path}")
|
||||||
|
|
||||||
|
ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
|
||||||
|
|
||||||
|
if not os.path.isdir(ds_checkpoint_dir):
|
||||||
|
raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
|
||||||
|
|
||||||
|
state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
|
||||||
|
if lazy_mode:
|
||||||
|
return state_dict
|
||||||
|
else:
|
||||||
|
return to_torch_tensor(state_dict)
|
||||||
|
|
||||||
|
|
||||||
|
def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
|
||||||
|
output_dir,
|
||||||
|
max_shard_size="5GB",
|
||||||
|
safe_serialization=False,
|
||||||
|
tag=None,
|
||||||
|
exclude_frozen_parameters=False):
|
||||||
|
"""
|
||||||
|
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
|
||||||
|
loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
|
||||||
|
|
||||||
|
Args:
|
||||||
|
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
|
||||||
|
- ``output_dir``: directory to the pytorch fp32 state_dict output files
|
||||||
|
- ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
|
||||||
|
- ``safe_serialization``: whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
|
||||||
|
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
|
||||||
|
- ``exclude_frozen_parameters``: exclude frozen parameters
|
||||||
|
"""
|
||||||
|
|
||||||
|
# Dependency pre-check
|
||||||
|
if safe_serialization:
|
||||||
|
try:
|
||||||
|
from safetensors.torch import save_file
|
||||||
|
except ImportError:
|
||||||
|
print('If you want to use `safe_serialization`, please `pip install safetensors`')
|
||||||
|
raise
|
||||||
|
if max_shard_size is not None:
|
||||||
|
try:
|
||||||
|
from huggingface_hub import split_torch_state_dict_into_shards
|
||||||
|
except ImportError:
|
||||||
|
print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
|
||||||
|
raise
|
||||||
|
|
||||||
|
# Convert zero checkpoint to state_dict
|
||||||
|
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
|
||||||
|
tag,
|
||||||
|
exclude_frozen_parameters,
|
||||||
|
lazy_mode=True)
|
||||||
|
|
||||||
|
# Shard the model if it is too big.
|
||||||
|
weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
|
||||||
|
if max_shard_size is not None:
|
||||||
|
filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
|
||||||
|
# an memory-efficient approach for sharding
|
||||||
|
empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
|
||||||
|
state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
|
||||||
|
filename_pattern=filename_pattern,
|
||||||
|
max_shard_size=max_shard_size)
|
||||||
|
else:
|
||||||
|
from collections import namedtuple
|
||||||
|
StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
|
||||||
|
state_dict_split = StateDictSplit(is_sharded=False,
|
||||||
|
filename_to_tensors={weights_name: list(state_dict.keys())})
|
||||||
|
|
||||||
|
# Save the model by shard
|
||||||
|
os.makedirs(output_dir, exist_ok=True)
|
||||||
|
filename_to_tensors = state_dict_split.filename_to_tensors.items()
|
||||||
|
for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
|
||||||
|
shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
|
||||||
|
shard_state_dict = to_torch_tensor(shard_state_dict)
|
||||||
|
output_path = os.path.join(output_dir, shard_file)
|
||||||
|
if safe_serialization:
|
||||||
|
save_file(shard_state_dict, output_path, metadata={"format": "pt"})
|
||||||
|
else:
|
||||||
|
torch.save(shard_state_dict, output_path)
|
||||||
|
# release the memory of current shard
|
||||||
|
for tensor_name in list(shard_state_dict.keys()):
|
||||||
|
del state_dict[tensor_name]
|
||||||
|
del shard_state_dict[tensor_name]
|
||||||
|
del shard_state_dict
|
||||||
|
gc.collect()
|
||||||
|
|
||||||
|
# Save index if sharded
|
||||||
|
if state_dict_split.is_sharded:
|
||||||
|
index = {
|
||||||
|
"metadata": state_dict_split.metadata,
|
||||||
|
"weight_map": state_dict_split.tensor_to_filename,
|
||||||
|
}
|
||||||
|
save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
|
||||||
|
save_index_file = os.path.join(output_dir, save_index_file)
|
||||||
|
with open(save_index_file, "w", encoding="utf-8") as f:
|
||||||
|
content = json.dumps(index, indent=2, sort_keys=True) + "\n"
|
||||||
|
f.write(content)
|
||||||
|
|
||||||
|
|
||||||
|
def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
|
||||||
|
"""
|
||||||
|
1. Put the provided model to cpu
|
||||||
|
2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
|
||||||
|
3. Load it into the provided model
|
||||||
|
|
||||||
|
Args:
|
||||||
|
- ``model``: the model object to update
|
||||||
|
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
|
||||||
|
- ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
|
||||||
|
|
||||||
|
Returns:
|
||||||
|
- ``model`: modified model
|
||||||
|
|
||||||
|
Make sure you have plenty of CPU memory available before you call this function. If you don't
|
||||||
|
have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
|
||||||
|
conveniently placed for you in the checkpoint folder.
|
||||||
|
|
||||||
|
A typical usage might be ::
|
||||||
|
|
||||||
|
from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
|
||||||
|
model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
|
||||||
|
# submit to model hub or save the model to share with others
|
||||||
|
|
||||||
|
Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
|
||||||
|
of the same application. i.e. you will need to re-initialize the deepspeed engine, since
|
||||||
|
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
|
||||||
|
|
||||||
|
"""
|
||||||
|
logger.info(f"Extracting fp32 weights")
|
||||||
|
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
|
||||||
|
|
||||||
|
logger.info(f"Overwriting model with fp32 weights")
|
||||||
|
model = model.cpu()
|
||||||
|
model.load_state_dict(state_dict, strict=False)
|
||||||
|
|
||||||
|
return model
|
||||||
|
|
||||||
|
|
||||||
|
if __name__ == "__main__":
|
||||||
|
parser = argparse.ArgumentParser()
|
||||||
|
parser.add_argument("checkpoint_dir",
|
||||||
|
type=str,
|
||||||
|
help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
|
||||||
|
parser.add_argument("output_dir",
|
||||||
|
type=str,
|
||||||
|
help="directory to the pytorch fp32 state_dict output files"
|
||||||
|
"(e.g. path/checkpoint-12-output/)")
|
||||||
|
parser.add_argument(
|
||||||
|
"--max_shard_size",
|
||||||
|
type=str,
|
||||||
|
default="5GB",
|
||||||
|
help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
|
||||||
|
"lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
|
||||||
|
"We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
|
||||||
|
"without CPU OOM issues.")
|
||||||
|
parser.add_argument(
|
||||||
|
"--safe_serialization",
|
||||||
|
default=False,
|
||||||
|
action='store_true',
|
||||||
|
help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
|
||||||
|
parser.add_argument("-t",
|
||||||
|
"--tag",
|
||||||
|
type=str,
|
||||||
|
default=None,
|
||||||
|
help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
|
||||||
|
parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
|
||||||
|
parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
|
||||||
|
args = parser.parse_args()
|
||||||
|
|
||||||
|
debug = args.debug
|
||||||
|
|
||||||
|
convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
|
||||||
|
args.output_dir,
|
||||||
|
max_shard_size=args.max_shard_size,
|
||||||
|
safe_serialization=args.safe_serialization,
|
||||||
|
tag=args.tag,
|
||||||
|
exclude_frozen_parameters=args.exclude_frozen_parameters)
|
||||||
Reference in New Issue
Block a user