初始化项目，由ModelHub XC社区提供模型

Model: flax-community/gpt-2-spanish
Source: Original Platform

.gitattributes

@@ -0,0 +1,18 @@
+*.bin.* filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tar.gz filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack 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
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+model.safetensors filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,43 @@
+---
+language: es
+tags:
+- text-generation
+datasets:
+- oscar
+widgets:
+- text: 'Érase un vez '
+- text: >-
+    Frase: Esta película es muy agradable. Sentimiento: positivo Frase: Odiaba
+    esta película, apesta. Sentimiento: negativo Frase: Esta película fue
+    bastante mala. Sentimiento: 
+license: apache-2.0
+---
+
+# Spanish GPT-2
+
+GPT-2 model trained from scratch on the Spanish portion of [OSCAR](https://huggingface.co/datasets/viewer/?dataset=oscar).
+The model is trained with Flax and using TPUs sponsored by Google since this is part of the
+[Flax/Jax Community Week](https://discuss.huggingface.co/t/open-to-the-community-community-week-using-jax-flax-for-nlp-cv/7104)
+organised by HuggingFace.
+
+## Model description
+
+The model used for training is [OpenAI's GPT-2](https://openai.com/blog/better-language-models/), introduced in the paper ["Language Models are Unsupervised Multitask Learners"](https://d4mucfpksywv.cloudfront.net/better-language-models/language_models_are_unsupervised_multitask_learners.pdf) by Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei and Ilya Sutskever.
+
+This model is available in the 🤗 [Model Hub](https://huggingface.co/gpt2).
+
+## Training data
+
+Spanish portion of OSCAR or **O**pen **S**uper-large **C**rawled **A**LMAnaCH co**R**pus, a huge multilingual corpus obtained by language classification and filtering of the [Common Crawl](https://commoncrawl.org/) corpus using the [goclassy](https://github.com/pjox/goclassy) architecture.
+
+This corpus is available in the 🤗 [Datasets](https://huggingface.co/datasets/oscar) library.
+
+## Team members
+- Manuel Romero ([mrm8488](https://huggingface.co/mrm8488))
+- María Grandury ([mariagrandury](https://huggingface.co/mariagrandury))
+- Pablo González de Prado ([Pablogps](https://huggingface.co/Pablogps))
+- Daniel Vera ([daveni](https://huggingface.co/daveni))
+- Sri Lakshmi ([srisweet](https://huggingface.co/srisweet))
+- José Posada ([jdposa](https://huggingface.co/jdposa))
+- Santiago Hincapie ([shpotes](https://huggingface.co/shpotes))
+- Jorge ([jorgealro](https://huggingface.co/jorgealro))

config.json

@@ -0,0 +1,36 @@
+{
+  "activation_function": "gelu_new",
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.0,
+  "bos_token_id": 50256,
+  "embd_pdrop": 0.0,
+  "eos_token_id": 50256,
+  "gradient_checkpointing": false,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 768,
+  "n_head": 12,
+  "n_inner": null,
+  "n_layer": 12,
+  "n_positions": 1024,
+  "resid_pdrop": 0.0,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "task_specific_params": {
+    "text-generation": {
+      "do_sample": true,
+      "max_length": 50
+    }
+  },
+  "transformers_version": "4.9.0.dev0",
+  "use_cache": true,
+  "vocab_size": 50257
+}

events.out.tfevents.1626191863.t1v-n-cf89aecf-w-0.861739.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f5a6cef1983c8dcbf34fc6d98bc192078bb1949efeea81672ee0edcf125fbd7
+size 367772

events.out.tfevents.1626201019.t1v-n-cf89aecf-w-0.880320.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0c3cf7a5e63d4a77d3e7d14b72a3f38a9e604efcb0e78faa3ece56e96707932
+size 735617

events.out.tfevents.1626210976.t1v-n-cf89aecf-w-0.890224.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:116109fbd23b28c6a4dc9ea828729181901502ced8d7c20ce220b9438ddc4341
+size 73496

events.out.tfevents.1626212077.t1v-n-cf89aecf-w-0.892752.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8d395a87948d65cb9d66207f2321ec9b3cfa06afe2d800d2b246fb3aa6385eb
+size 11003489

events.out.tfevents.1626343565.t1v-n-cf89aecf-w-0.1005830.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a13db211fb1e8fcab57645f6987bab32b307ade0e149d5954194e5e433200c12
+size 1397738

events.out.tfevents.1626354152.t1v-n-cf89aecf-w-0.1016609.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0224f764bf692ac3bd6d6b158147621d510be7926328ca69c4b91c71846e0442
+size 1397738

events.out.tfevents.1626364173.t1v-n-cf89aecf-w-0.1026566.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49d618c3b6ece936a9e3ae7cfcf162166289a2ad3729a48abf2dd46ac0728b02
+size 40

events.out.tfevents.1626364345.t1v-n-cf89aecf-w-0.1028406.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee72a372d2f2bbe5ab9a307269cabc0e6d36909d6096ad5dbb58c546c57ad9d5
+size 11059767

events.out.tfevents.1626437960.t1v-n-cf89aecf-w-0.1093227.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e852c812c8c31772be718e7995671cc231b73689899ca307c67d834e1537a51
+size 14749

events.out.tfevents.1626440707.t1v-n-cf89aecf-w-0.1097198.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d91156cf42fec78270c1a1bad4c3e771c71ddc50987c79fc26604691ed4ac2bc
+size 9293307

flax_model.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5e3bd9282196a12dd5de77e69d8b231f6077e9dbe237586ad6d349fa544e281
+size 497764120

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31ce24d2db59bcc5a7b715591334f4762d4ea58e3efb595d126c80a94cff8cd7
+size 510359598

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e56fb07dd88794a4374ebafa12178489cfeb4a1f029d2ed0337bc3c1268d71ca
+size 510401385

run.sh

@@ -0,0 +1,23 @@
+./run_clm_flax.py \
+    --output_dir="./" \
+    --model_type="gpt2" \
+    --model_name_or_path="./" \
+    --config_name="./" \
+    --tokenizer_name="./" \
+    --dataset_name="oscar" \
+    --dataset_config_name="unshuffled_deduplicated_es" \
+    --do_train \
+    --do_eval \
+    --block_size="512" \
+    --per_device_train_batch_size="64" \
+    --per_device_eval_batch_size="64" \
+    --adafactor \
+    --learning_rate="3e-3" --warmup_steps="1000" \
+    --adam_beta1="0.9" --adam_beta2="0.98" --weight_decay="0.001" \
+    --overwrite_output_dir \
+    --num_train_epochs="20" \
+    --logging_steps="100" \
+    --save_steps="2500" \
+    --eval_steps="1000000" \
+    --preprocessing_num_workers="64"
+

run_clm_flax.py

@@ -0,0 +1 @@
+../../examples/flax/language-modeling/run_clm_flax.py

train_tokenizer.py

@@ -0,0 +1,20 @@
+#!/usr/bin/env python3
+from datasets import load_dataset
+from tokenizers import ByteLevelBPETokenizer
+# load dataset
+dataset = load_dataset("oscar", "unshuffled_deduplicated_es", split="train")
+# Instantiate tokenizer
+tokenizer = ByteLevelBPETokenizer()
+def batch_iterator(batch_size=10000):
+    for i in range(0, len(dataset), batch_size):
+        yield dataset[i: i + batch_size]["text"]
+# Customized training
+tokenizer.train_from_iterator(batch_iterator(), vocab_size=50265, min_frequency=2, special_tokens=[
+    "<s>",
+    "<pad>",
+    "</s>",
+    "<unk>",
+    "<mask>",
+])
+# Save files to disk
+tokenizer.save("./tokenizer.json")