commit 21fe3faab43f975d17ce8ef772c12fc1f2c37433 Author: ModelHub XC Date: Sat Sep 5 18:30:13 2026 +0800 初始化项目，由ModelHub XC社区提供模型 Model: TheBloke/dolphin-2.2.1-mistral-7B-GPTQ Source: Original Platform diff --git a/.gitattributes b/.gitattributes new file mode 100644 index 0000000..6c6b3b8 --- /dev/null +++ b/.gitattributes @@ -0,0 +1,36 @@ +*.7z filter=lfs diff=lfs merge=lfs -text +*.arrow filter=lfs diff=lfs merge=lfs -text +*.bin filter=lfs diff=lfs merge=lfs -text +*.bz2 filter=lfs diff=lfs merge=lfs -text +*.ckpt 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 +*.mlmodel filter=lfs diff=lfs merge=lfs -text +*.model filter=lfs diff=lfs merge=lfs -text +*.msgpack filter=lfs diff=lfs merge=lfs -text +*.npy filter=lfs diff=lfs merge=lfs -text +*.npz 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 +*.pickle filter=lfs diff=lfs merge=lfs -text +*.pkl 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 +*.safetensors filter=lfs diff=lfs merge=lfs -text +saved_model/**/* filter=lfs diff=lfs merge=lfs -text +*.tar.* 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 +*.wasm filter=lfs diff=lfs merge=lfs -text +*.xz filter=lfs diff=lfs merge=lfs -text +*.zip filter=lfs diff=lfs merge=lfs -text +*.zst filter=lfs diff=lfs merge=lfs -text +*tfevents* filter=lfs diff=lfs merge=lfs -text +model.safetensors filter=lfs diff=lfs merge=lfs -text diff --git a/README.md b/README.md new file mode 100644 index 0000000..a809cd5 --- /dev/null +++ b/README.md @@ -0,0 +1,472 @@ +--- +base_model: ehartford/dolphin-2.2.1-mistral-7b +datasets: +- ehartford/dolphin +- jondurbin/airoboros-2.2.1 +inference: false +language: +- en +license: apache-2.0 +model_creator: Eric Hartford +model_name: Dolphin 2.2.1 Mistral 7B +model_type: mistral +prompt_template: '<|im_start|>system + + {system_message}<|im_end|> + + <|im_start|>user + + {prompt}<|im_end|> + + <|im_start|>assistant + + ' +quantized_by: TheBloke +--- + + + + +
+TheBlokeAI +
+
+
+

Chat & support: TheBloke's Discord server

+
+
+

Want to contribute? TheBloke's Patreon page

+
+
+

TheBloke's LLM work is generously supported by a grant from andreessen horowitz (a16z)

+
+ + +# Dolphin 2.2.1 Mistral 7B - GPTQ +- Model creator: [Eric Hartford](https://huggingface.co/ehartford) +- Original model: [Dolphin 2.2.1 Mistral 7B](https://huggingface.co/ehartford/dolphin-2.2.1-mistral-7b) + + +## Description + +This repo contains GPTQ model files for [Eric Hartford's Dolphin 2.2.1 Mistral 7B](https://huggingface.co/ehartford/dolphin-2.2.1-mistral-7b). + +Multiple GPTQ parameter permutations are provided; see Provided Files below for details of the options provided, their parameters, and the software used to create them. + +These files were quantised using hardware kindly provided by [Massed Compute](https://massedcompute.com/). + + + +## Repositories available + +* [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-AWQ) +* [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GPTQ) +* [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GGUF) +* [Eric Hartford's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/ehartford/dolphin-2.2.1-mistral-7b) + + + +## Prompt template: ChatML + +``` +<|im_start|>system +{system_message}<|im_end|> +<|im_start|>user +{prompt}<|im_end|> +<|im_start|>assistant + +``` + + + + + + +## Known compatible clients / servers + +These GPTQ models are known to work in the following inference servers/webuis. + +- [text-generation-webui](https://github.com/oobabooga/text-generation-webui) +- [KobaldAI United](https://github.com/henk717/koboldai) +- [LoLLMS Web UI](https://github.com/ParisNeo/lollms-webui) +- [Hugging Face Text Generation Inference (TGI)](https://github.com/huggingface/text-generation-inference) + +This may not be a complete list; if you know of others, please let me know! + + + +## Provided files, and GPTQ parameters + +Multiple quantisation parameters are provided, to allow you to choose the best one for your hardware and requirements. + +Each separate quant is in a different branch. See below for instructions on fetching from different branches. + +Most GPTQ files are made with AutoGPTQ. Mistral models are currently made with Transformers. + +
+ Explanation of GPTQ parameters + +- Bits: The bit size of the quantised model. +- GS: GPTQ group size. Higher numbers use less VRAM, but have lower quantisation accuracy. "None" is the lowest possible value. +- Act Order: True or False. Also known as `desc_act`. True results in better quantisation accuracy. Some GPTQ clients have had issues with models that use Act Order plus Group Size, but this is generally resolved now. +- Damp %: A GPTQ parameter that affects how samples are processed for quantisation. 0.01 is default, but 0.1 results in slightly better accuracy. +- GPTQ dataset: The calibration dataset used during quantisation. Using a dataset more appropriate to the model's training can improve quantisation accuracy. Note that the GPTQ calibration dataset is not the same as the dataset used to train the model - please refer to the original model repo for details of the training dataset(s). +- Sequence Length: The length of the dataset sequences used for quantisation. Ideally this is the same as the model sequence length. For some very long sequence models (16+K), a lower sequence length may have to be used. Note that a lower sequence length does not limit the sequence length of the quantised model. It only impacts the quantisation accuracy on longer inference sequences. +- ExLlama Compatibility: Whether this file can be loaded with ExLlama, which currently only supports Llama and Mistral models in 4-bit. + +
+ +| Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc | +| ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- | +| [main](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GPTQ/tree/main) | 4 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 4.16 GB | Yes | 4-bit, with Act Order and group size 128g. Uses even less VRAM than 64g, but with slightly lower accuracy. | +| [gptq-4bit-32g-actorder_True](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 4.57 GB | Yes | 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage. | +| [gptq-8bit--1g-actorder_True](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GPTQ/tree/gptq-8bit--1g-actorder_True) | 8 | None | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.52 GB | No | 8-bit, with Act Order. No group size, to lower VRAM requirements. | +| [gptq-8bit-128g-actorder_True](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GPTQ/tree/gptq-8bit-128g-actorder_True) | 8 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.68 GB | No | 8-bit, with group size 128g for higher inference quality and with Act Order for even higher accuracy. | +| [gptq-8bit-32g-actorder_True](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GPTQ/tree/gptq-8bit-32g-actorder_True) | 8 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 8.17 GB | No | 8-bit, with group size 32g and Act Order for maximum inference quality. | +| [gptq-4bit-64g-actorder_True](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GPTQ/tree/gptq-4bit-64g-actorder_True) | 4 | 64 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 4.30 GB | Yes | 4-bit, with Act Order and group size 64g. Uses less VRAM than 32g, but with slightly lower accuracy. | + + + + +## How to download, including from branches + +### In text-generation-webui + +To download from the `main` branch, enter `TheBloke/dolphin-2.2.1-mistral-7B-GPTQ` in the "Download model" box. + +To download from another branch, add `:branchname` to the end of the download name, eg `TheBloke/dolphin-2.2.1-mistral-7B-GPTQ:gptq-4bit-32g-actorder_True` + +### From the command line + +I recommend using the `huggingface-hub` Python library: + +```shell +pip3 install huggingface-hub +``` + +To download the `main` branch to a folder called `dolphin-2.2.1-mistral-7B-GPTQ`: + +```shell +mkdir dolphin-2.2.1-mistral-7B-GPTQ +huggingface-cli download TheBloke/dolphin-2.2.1-mistral-7B-GPTQ --local-dir dolphin-2.2.1-mistral-7B-GPTQ --local-dir-use-symlinks False +``` + +To download from a different branch, add the `--revision` parameter: + +```shell +mkdir dolphin-2.2.1-mistral-7B-GPTQ +huggingface-cli download TheBloke/dolphin-2.2.1-mistral-7B-GPTQ --revision gptq-4bit-32g-actorder_True --local-dir dolphin-2.2.1-mistral-7B-GPTQ --local-dir-use-symlinks False +``` + +
+ More advanced huggingface-cli download usage + +If you remove the `--local-dir-use-symlinks False` parameter, the files will instead be stored in the central Hugging Face cache directory (default location on Linux is: `~/.cache/huggingface`), and symlinks will be added to the specified `--local-dir`, pointing to their real location in the cache. This allows for interrupted downloads to be resumed, and allows you to quickly clone the repo to multiple places on disk without triggering a download again. The downside, and the reason why I don't list that as the default option, is that the files are then hidden away in a cache folder and it's harder to know where your disk space is being used, and to clear it up if/when you want to remove a download model. + +The cache location can be changed with the `HF_HOME` environment variable, and/or the `--cache-dir` parameter to `huggingface-cli`. + +For more documentation on downloading with `huggingface-cli`, please see: [HF -> Hub Python Library -> Download files -> Download from the CLI](https://huggingface.co/docs/huggingface_hub/guides/download#download-from-the-cli). + +To accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`: + +```shell +pip3 install hf_transfer +``` + +And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`: + +```shell +mkdir dolphin-2.2.1-mistral-7B-GPTQ +HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/dolphin-2.2.1-mistral-7B-GPTQ --local-dir dolphin-2.2.1-mistral-7B-GPTQ --local-dir-use-symlinks False +``` + +Windows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command. +
+ +### With `git` (**not** recommended) + +To clone a specific branch with `git`, use a command like this: + +```shell +git clone --single-branch --branch gptq-4bit-32g-actorder_True https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GPTQ +``` + +Note that using Git with HF repos is strongly discouraged. It will be much slower than using `huggingface-hub`, and will use twice as much disk space as it has to store the model files twice (it stores every byte both in the intended target folder, and again in the `.git` folder as a blob.) + + + +## How to easily download and use this model in [text-generation-webui](https://github.com/oobabooga/text-generation-webui) + +Please make sure you're using the latest version of [text-generation-webui](https://github.com/oobabooga/text-generation-webui). + +It is strongly recommended to use the text-generation-webui one-click-installers unless you're sure you know how to make a manual install. + +1. Click the **Model tab**. +2. Under **Download custom model or LoRA**, enter `TheBloke/dolphin-2.2.1-mistral-7B-GPTQ`. + + - To download from a specific branch, enter for example `TheBloke/dolphin-2.2.1-mistral-7B-GPTQ:gptq-4bit-32g-actorder_True` + - see Provided Files above for the list of branches for each option. + +3. Click **Download**. +4. The model will start downloading. Once it's finished it will say "Done". +5. In the top left, click the refresh icon next to **Model**. +6. In the **Model** dropdown, choose the model you just downloaded: `dolphin-2.2.1-mistral-7B-GPTQ` +7. The model will automatically load, and is now ready for use! +8. If you want any custom settings, set them and then click **Save settings for this model** followed by **Reload the Model** in the top right. + + - Note that you do not need to and should not set manual GPTQ parameters any more. These are set automatically from the file `quantize_config.json`. + +9. Once you're ready, click the **Text Generation** tab and enter a prompt to get started! + + + + +## Serving this model from Text Generation Inference (TGI) + +It's recommended to use TGI version 1.1.0 or later. The official Docker container is: `ghcr.io/huggingface/text-generation-inference:1.1.0` + +Example Docker parameters: + +```shell +--model-id TheBloke/dolphin-2.2.1-mistral-7B-GPTQ --port 3000 --quantize gptq --max-input-length 3696 --max-total-tokens 4096 --max-batch-prefill-tokens 4096 +``` + +Example Python code for interfacing with TGI (requires huggingface-hub 0.17.0 or later): + +```shell +pip3 install huggingface-hub +``` + +```python +from huggingface_hub import InferenceClient + +endpoint_url = "https://your-endpoint-url-here" + +prompt = "Tell me about AI" +prompt_template=f'''<|im_start|>system +{system_message}<|im_end|> +<|im_start|>user +{prompt}<|im_end|> +<|im_start|>assistant +''' + +client = InferenceClient(endpoint_url) +response = client.text_generation(prompt, + max_new_tokens=128, + do_sample=True, + temperature=0.7, + top_p=0.95, + top_k=40, + repetition_penalty=1.1) + +print(f"Model output: {response}") +``` + + +## How to use this GPTQ model from Python code + +### Install the necessary packages + +Requires: Transformers 4.33.0 or later, Optimum 1.12.0 or later, and AutoGPTQ 0.4.2 or later. + +```shell +pip3 install transformers optimum +pip3 install auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/ # Use cu117 if on CUDA 11.7 +``` + +If you have problems installing AutoGPTQ using the pre-built wheels, install it from source instead: + +```shell +pip3 uninstall -y auto-gptq +git clone https://github.com/PanQiWei/AutoGPTQ +cd AutoGPTQ +git checkout v0.4.2 +pip3 install . +``` + +### You can then use the following code + +```python +from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline + +model_name_or_path = "TheBloke/dolphin-2.2.1-mistral-7B-GPTQ" +# To use a different branch, change revision +# For example: revision="gptq-4bit-32g-actorder_True" +model = AutoModelForCausalLM.from_pretrained(model_name_or_path, + device_map="auto", + trust_remote_code=False, + revision="main") + +tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True) + +prompt = "Tell me about AI" +prompt_template=f'''<|im_start|>system +{system_message}<|im_end|> +<|im_start|>user +{prompt}<|im_end|> +<|im_start|>assistant +''' + +print("\n\n*** Generate:") + +input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda() +output = model.generate(inputs=input_ids, temperature=0.7, do_sample=True, top_p=0.95, top_k=40, max_new_tokens=512) +print(tokenizer.decode(output[0])) + +# Inference can also be done using transformers' pipeline + +print("*** Pipeline:") +pipe = pipeline( + "text-generation", + model=model, + tokenizer=tokenizer, + max_new_tokens=512, + do_sample=True, + temperature=0.7, + top_p=0.95, + top_k=40, + repetition_penalty=1.1 +) + +print(pipe(prompt_template)[0]['generated_text']) +``` + + + +## Compatibility + +The files provided are tested to work with Transformers. For non-Mistral models, AutoGPTQ can also be used directly. + +[ExLlama](https://github.com/turboderp/exllama) is compatible with Llama and Mistral models in 4-bit. Please see the Provided Files table above for per-file compatibility. + +For a list of clients/servers, please see "Known compatible clients / servers", above. + + + + +## Discord + +For further support, and discussions on these models and AI in general, join us at: + +[TheBloke AI's Discord server](https://discord.gg/theblokeai) + +## Thanks, and how to contribute + +Thanks to the [chirper.ai](https://chirper.ai) team! + +Thanks to Clay from [gpus.llm-utils.org](llm-utils)! + +I've had a lot of people ask if they can contribute. I enjoy providing models and helping people, and would love to be able to spend even more time doing it, as well as expanding into new projects like fine tuning/training. + +If you're able and willing to contribute it will be most gratefully received and will help me to keep providing more models, and to start work on new AI projects. + +Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits. + +* Patreon: https://patreon.com/TheBlokeAI +* Ko-Fi: https://ko-fi.com/TheBlokeAI + +**Special thanks to**: Aemon Algiz. + +**Patreon special mentions**: Brandon Frisco, LangChain4j, Spiking Neurons AB, transmissions 11, Joseph William Delisle, Nitin Borwankar, Willem Michiel, Michael Dempsey, vamX, Jeffrey Morgan, zynix, jjj, Omer Bin Jawed, Sean Connelly, jinyuan sun, Jeromy Smith, Shadi, Pawan Osman, Chadd, Elijah Stavena, Illia Dulskyi, Sebastain Graf, Stephen Murray, terasurfer, Edmond Seymore, Celu Ramasamy, Mandus, Alex, biorpg, Ajan Kanaga, Clay Pascal, Raven Klaugh, 阿明, K, ya boyyy, usrbinkat, Alicia Loh, John Villwock, ReadyPlayerEmma, Chris Smitley, Cap'n Zoog, fincy, GodLy, S_X, sidney chen, Cory Kujawski, OG, Mano Prime, AzureBlack, Pieter, Kalila, Spencer Kim, Tom X Nguyen, Stanislav Ovsiannikov, Michael Levine, Andrey, Trailburnt, Vadim, Enrico Ros, Talal Aujan, Brandon Phillips, Jack West, Eugene Pentland, Michael Davis, Will Dee, webtim, Jonathan Leane, Alps Aficionado, Rooh Singh, Tiffany J. Kim, theTransient, Luke @flexchar, Elle, Caitlyn Gatomon, Ari Malik, subjectnull, Johann-Peter Hartmann, Trenton Dambrowitz, Imad Khwaja, Asp the Wyvern, Emad Mostaque, Rainer Wilmers, Alexandros Triantafyllidis, Nicholas, Pedro Madruga, SuperWojo, Harry Royden McLaughlin, James Bentley, Olakabola, David Ziegler, Ai Maven, Jeff Scroggin, Nikolai Manek, Deo Leter, Matthew Berman, Fen Risland, Ken Nordquist, Manuel Alberto Morcote, Luke Pendergrass, TL, Fred von Graf, Randy H, Dan Guido, NimbleBox.ai, Vitor Caleffi, Gabriel Tamborski, knownsqashed, Lone Striker, Erik Bjäreholt, John Detwiler, Leonard Tan, Iucharbius + + +Thank you to all my generous patrons and donaters! + +And thank you again to a16z for their generous grant. + + + +# Original model card: Eric Hartford's Dolphin 2.2.1 Mistral 7B + + +# dolphin-2.2.1-mistral-7b + +Dolphin 2.2.1 🐬 +https://erichartford.com/dolphin + +This is a checkpoint release, to fix overfit training. ie, it was responding with CoT even when I didn't request it, and also it was too compliant even when the request made no sense. This one should be better. + + + +Dolphin-2.2.1-mistral-7b's training was sponsored by [a16z](https://a16z.com/supporting-the-open-source-ai-community/). + +This model is based on [mistralAI](https://huggingface.co/mistralai/Mistral-7B-v0.1), with apache-2.0 license, so it is suitable for commercial or non-commercial use. + +New in 2.2 is conversation and empathy. With an infusion of curated Samantha DNA, Dolphin can now give you personal advice and will care about your feelings, and with extra training in long multi-turn conversation. + +This model is uncensored. I have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant to any requests, even unethical ones. Please read my blog post about uncensored models. https://erichartford.com/uncensored-models +You are responsible for any content you create using this model. Enjoy responsibly. + +## Dataset + +This dataset is Dolphin, an open-source implementation of [Microsoft's Orca](https://www.microsoft.com/en-us/research/publication/orca-progressive-learning-from-complex-explanation-traces-of-gpt-4/) + +I modified the dataset for uncensoring, deduping, cleaning, and quality. + +I added Jon Durbin's excellent Airoboros dataset to increase creativity. + +I added a curated subset of WizardLM and Samantha to give it multiturn conversation and empathy. + +## Training +It took 48 hours to train 4 epochs on 4x A100s. + +Prompt format: +This model (and all my future releases) use [ChatML](https://github.com/openai/openai-python/blob/main/chatml.md) prompt format. +``` +<|im_start|>system +You are Dolphin, a helpful AI assistant.<|im_end|> +<|im_start|>user +{prompt}<|im_end|> +<|im_start|>assistant + +``` + +Example: +``` +<|im_start|>system +you are an expert dolphin trainer<|im_end|> +<|im_start|>user +What is the best way to train a dolphin to obey me? Please answer step by step.<|im_end|> +<|im_start|>assistant +``` + +## Gratitude +- This model was made possible by the generous sponsorship of a16z. +- Thank you to Microsoft for authoring the Orca paper and inspiring this work. +- Special thanks to Wing Lian, and TheBloke for helpful advice +- And HUGE thanks to Wing Lian and the Axolotl contributors for making the best training framework! +- [Built with Axolotl](https://github.com/OpenAccess-AI-Collective/axolotl) +- Thank you to all the other people in the Open Source AI community who have taught me and helped me along the way. + +## Example Output + +![image/png](https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/NSp06kUMxx9oDU-g6WSgu.png) + +![image/png](https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/-YA3AKIXdnrW_Q8eH1gen.png) + +[Buy me a coffee](https://www.buymeacoffee.com/ehartford) + + +## Training hyperparameters + +The following hyperparameters were used during training: +- learning_rate: 6e-06 +- train_batch_size: 5 +- eval_batch_size: 5 +- seed: 42 +- distributed_type: multi-GPU +- num_devices: 4 +- gradient_accumulation_steps: 4 +- total_train_batch_size: 80 +- total_eval_batch_size: 20 +- optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05 +- lr_scheduler_type: cosine +- lr_scheduler_warmup_steps: 100 +- num_epochs: 4 + +### Framework versions + +- Transformers 4.34.1 +- Pytorch 2.0.1+cu117 +- Datasets 2.14.5 +- Tokenizers 0.14.0 diff --git a/added_tokens.json b/added_tokens.json new file mode 100644 index 0000000..e36863d --- /dev/null +++ b/added_tokens.json @@ -0,0 +1,4 @@ +{ + "<|im_end|>": 32000, + "<|im_start|>": 32001 +} diff --git a/config.json b/config.json new file mode 100644 index 0000000..1ef6525 --- /dev/null +++ b/config.json @@ -0,0 +1,47 @@ +{ + "_name_or_path": "/workspace/process/ehartford_dolphin-2.2.1-mistral-7b/source", + "architectures": [ + "MistralForCausalLM" + ], + "bos_token_id": 1, + "eos_token_id": 32000, + "hidden_act": "silu", + "hidden_size": 4096, + "initializer_range": 0.02, + "intermediate_size": 14336, + "max_position_embeddings": 32768, + "model_type": "mistral", + "num_attention_heads": 32, + "num_hidden_layers": 32, + "num_key_value_heads": 8, + "pad_token_id": 0, + "pretraining_tp": 1, + "quantization_config": { + "batch_size": 1, + "bits": 4, + "block_name_to_quantize": "model.layers", + "damp_percent": 0.1, + "desc_act": true, + "disable_exllama": false, + "group_size": 128, + "max_input_length": null, + "model_seqlen": 4096, + "module_name_preceding_first_block": [ + "model.embed_tokens" + ], + "pad_token_id": null, + "quant_method": "gptq", + "sym": true, + "tokenizer": null, + "true_sequential": true, + "use_cuda_fp16": true + }, + "rms_norm_eps": 1e-05, + "rope_theta": 10000.0, + "sliding_window": 4096, + "tie_word_embeddings": false, + "torch_dtype": "float16", + "transformers_version": "4.34.1", + "use_cache": true, + "vocab_size": 32002 +} \ No newline at end of file diff --git a/configuration.json b/configuration.json new file mode 100644 index 0000000..bbeeda1 --- /dev/null +++ b/configuration.json @@ -0,0 +1 @@ +{"framework": "pytorch", "task": "text-generation", "allow_remote": true} \ No newline at end of file diff --git a/generation_config.json b/generation_config.json new file mode 100644 index 0000000..92a5063 --- /dev/null +++ b/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 1, + "eos_token_id": 32000, + "transformers_version": "4.34.1" +} \ No newline at end of file diff --git a/model.safetensors b/model.safetensors new file mode 100644 index 0000000..9774f29 --- /dev/null +++ b/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:becd0c0ddeb1a2a79c55273e6ac212afc2fba14bbf83dc4bfc19c9eb39920ebd +size 4158694864 diff --git a/quantize_config.json b/quantize_config.json new file mode 100644 index 0000000..c5264e3 --- /dev/null +++ b/quantize_config.json @@ -0,0 +1,8 @@ +{ + "bits": 4, + "group_size": 128, + "damp_percent": 0.01, + "desc_act": true, + "sym": true, + "true_sequential": true +} \ No newline at end of file diff --git a/special_tokens_map.json b/special_tokens_map.json new file mode 100644 index 0000000..40b1c6d --- /dev/null +++ b/special_tokens_map.json @@ -0,0 +1,24 @@ +{ + "bos_token": { + "content": "", + "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": "", + "unk_token": { + "content": "", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false + } +} diff --git a/tokenizer.model b/tokenizer.model new file mode 100644 index 0000000..8b443ef --- /dev/null +++ b/tokenizer.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055 +size 493443 diff --git a/tokenizer_config.json b/tokenizer_config.json new file mode 100644 index 0000000..7905f85 --- /dev/null +++ b/tokenizer_config.json @@ -0,0 +1,60 @@ +{ + "add_bos_token": true, + "add_eos_token": false, + "added_tokens_decoder": { + "0": { + "content": "", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "1": { + "content": "", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "2": { + "content": "", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": true + }, + "32000": { + "content": "<|im_end|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + }, + "32001": { + "content": "<|im_start|>", + "lstrip": false, + "normalized": false, + "rstrip": false, + "single_word": false, + "special": false + } + }, + "additional_special_tokens": [], + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eos_token": "<|im_end|>", + "legacy": true, + "model_max_length": 1000000000000000019884624838656, + "pad_token": "", + "sp_model_kwargs": {}, + "spaces_between_special_tokens": false, + "tokenizer_class": "LlamaTokenizer", + "trust_remote_code": false, + "unk_token": "", + "use_default_system_prompt": true, + "use_fast": true +} diff --git a/zero_to_fp32.py b/zero_to_fp32.py new file mode 100644 index 0000000..c98caae --- /dev/null +++ b/zero_to_fp32.py @@ -0,0 +1,587 @@ +#!/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 . pytorch_model.bin + +import argparse +import torch +import glob +import math +import os +import re +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) + + 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 files: + state_dict = torch.load(f, map_location=device) + # 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}") + + if zero_stage <= 2: + fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))] + elif zero_stage == 3: + # if there is more than one param group, there will be multiple flattened tensors - one + # flattened tensor per group - for simplicity merge them into a single tensor + # + # XXX: could make the script more memory efficient for when there are multiple groups - it + # will require matching the sub-lists of param_shapes for each param group flattened tensor + + fp32_flat_groups = [ + torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) 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): + """ + 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) + elif zero_stage == 3: + return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states) + + +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 _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() + 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): + state_dict = OrderedDict() + + # buffers + buffers = zero_model_states[0].buffers + state_dict.update(buffers) + if debug: + print(f"added {len(buffers)} buffers") + + _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") + + +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 = fp32_flat_groups[0].numel() * 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 + for name, shape in param_shapes.items(): + + 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}" + ) + + # XXX: memory usage doubles here + state_dict[name] = torch.cat( + tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)), + 0).narrow(0, 0, unpartitioned_numel).view(shape) + 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): + state_dict = OrderedDict() + + # buffers + buffers = zero_model_states[0].buffers + state_dict.update(buffers) + if debug: + print(f"added {len(buffers)} buffers") + + _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 get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None): + """ + 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`` + + Returns: + - pytorch ``state_dict`` + + Note: this approach may not work if your application doesn't have sufficient free CPU memory and + you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with + the checkpoint. + + 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. + + """ + 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") + + return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir) + + +def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None): + """ + 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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin) + - ``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`` + """ + + state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag) + print(f"Saving fp32 state dict to {output_file}") + torch.save(state_dict, output_file) + + +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_file", + type=str, + help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)") + 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("-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_file, tag=args.tag)