prakharz
/

DIAL-FLANT5-XL

@@ -3,24 +3,29 @@ license: apache-2.0
 tags:
 - generated_from_trainer
 model-index:
-- name: t0-alltasksv2-t2
   results: []
 ---
-<!-- This model card has been generated automatically according to the information the Trainer had access to. You
-should probably proofread and complete it, then remove this comment. -->
-# t0-alltasksv2-t2
-This model is a fine-tuned version of [google/flan-t5-xl](https://huggingface.co/google/flan-t5-xl) on an unknown dataset.
-It achieves the following results on the evaluation set:
-- Loss: 1.2461
-- Train Runtime: 54501.5741
-- Train Samples Per Second: 17.607
-- Train Steps Per Second: 0.196
-- Train Loss: 1.2518
-- Train Samples: 239899
-- Gen Len: 9.0377
 ## Model description
@@ -32,7 +37,7 @@ More information needed
 ## Training and evaluation data
-More information needed
 ## Training procedure
@@ -40,53 +45,21 @@ More information needed
 The following hyperparameters were used during training:
 - learning_rate: 5e-05
-- train_batch_size: 3
-- eval_batch_size: 4
 - seed: 42
 - distributed_type: multi-GPU
-- num_devices: 6
-- gradient_accumulation_steps: 5
-- total_train_batch_size: 90
-- total_eval_batch_size: 24
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
-- num_epochs: 4.0
-### Training results
-| Training Loss | Epoch | Step  | Validation Loss | Rouge1  | Rouge2 | Rougel  | Rougelsum | Accuracy | F1      | Recall  | Precision | Bleu 1 | Bleu 2 | Bleu 3 | Bleu 4 | Rouge L | Gen Len |
-|:-------------:|:-----:|:-----:|:---------------:|:-------:|:------:|:-------:|:---------:|:--------:|:-------:|:-------:|:---------:|:------:|:------:|:------:|:------:|:-------:|:-------:|
-| 1.5099        | 0.15  | 400   | 1.3184          | 62.9956 | 6.6989 | 62.5464 | 62.7932   | 73.1148  | 73.1148 | 73.1148 | 73.1148   | 0.6509 | 0.0005 | 0.0001 | 0.0    | 0.6033  | 6.266   |
-| 1.4929        | 0.3   | 800   | 1.2843          | 64.2143 | 6.6289 | 63.7888 | 63.9982   | 75.1756  | 75.1756 | 75.1756 | 75.1756   | 0.6556 | 0.0005 | 0.0001 | 0.0    | 0.6128  | 6.5423  |
-| 1.4285        | 0.45  | 1200  | 1.2717          | 64.3279 | 6.63   | 63.9036 | 64.1136   | 75.4567  | 75.4567 | 75.4567 | 75.4567   | 0.6591 | 0.0005 | 0.0001 | 0.0    | 0.6153  | 6.4407  |
-| 1.4419        | 0.6   | 1600  | 1.2583          | 64.9421 | 6.5473 | 64.445  | 64.6707   | 76.1593  | 76.1593 | 76.1593 | 76.1593   | 0.6682 | 0.0005 | 0.0001 | 0.0    | 0.6218  | 6.378   |
-| 1.3967        | 0.75  | 2000  | 1.2497          | 65.9619 | 6.8233 | 65.4883 | 65.6707   | 77.2834  | 77.2834 | 77.2834 | 77.2834   | 0.6764 | 0.0005 | 0.0001 | 0.0    | 0.6297  | 6.4027  |
-| 1.4062        | 0.9   | 2400  | 1.2440          | 65.8936 | 6.6978 | 65.4631 | 65.6771   | 77.2365  | 77.2365 | 77.2365 | 77.2365   | 0.6753 | 0.0005 | 0.0001 | 0.0    | 0.6291  | 6.4097  |
-| 1.2708        | 1.05  | 2800  | 1.2460          | 66.4183 | 6.7096 | 66.0049 | 66.191    | 78.1265  | 78.1265 | 78.1265 | 78.1265   | 0.6781 | 0.0006 | 0.0001 | 0.0    | 0.6321  | 6.45    |
-| 1.2593        | 1.2   | 3200  | 1.2467          | 66.985  | 6.4882 | 66.596  | 66.7582   | 79.1569  | 79.1569 | 79.1569 | 79.1569   | 0.683  | 0.0006 | 0.0001 | 0.0    | 0.6395  | 6.495   |
-| 1.2623        | 1.35  | 3600  | 1.2461          | 66.8681 | 6.6985 | 66.4877 | 66.6731   | 78.7354  | 78.7354 | 78.7354 | 78.7354   | 0.6821 | 0.0006 | 0.0001 | 0.0    | 0.6369  | 6.4727  |
-| 1.2579        | 1.5   | 4000  | 1.2447          | 67.8146 | 6.7285 | 67.3078 | 67.5351   | 79.8126  | 79.8126 | 79.8126 | 79.8126   | 0.6937 | 0.0006 | 0.0001 | 0.0    | 0.6455  | 6.3997  |
-| 1.3159        | 1.65  | 4400  | 1.2281          | 67.8172 | 6.7857 | 67.3662 | 67.5871   | 79.9532  | 79.9532 | 79.9532 | 79.9532   | 0.694  | 0.0006 | 0.0001 | 0.0    | 0.6461  | 6.448   |
-| 1.2492        | 1.8   | 4800  | 1.2310          | 68.3986 | 6.8058 | 67.9201 | 68.1741   | 80.7963  | 80.7963 | 80.7963 | 80.7963   | 0.6991 | 0.0006 | 0.0001 | 0.0    | 0.6516  | 6.4757  |
-| 1.2338        | 1.95  | 5200  | 1.2253          | 67.9938 | 6.9092 | 67.5083 | 67.6913   | 80.0     | 80.0    | 80.0    | 80.0      | 0.695  | 0.0006 | 0.0001 | 0.0    | 0.6466  | 6.4163  |
-| 1.1788        | 2.1   | 5600  | 1.2499          | 67.9377 | 6.8965 | 67.4813 | 67.674    | 80.0937  | 80.0937 | 80.0937 | 80.0937   | 0.6917 | 0.0006 | 0.0001 | 0.0    | 0.6434  | 6.5223  |
-| 1.1833        | 2.25  | 6000  | 1.2401          | 68.1452 | 6.8628 | 67.7046 | 67.9493   | 80.3279  | 80.3279 | 80.3279 | 80.3279   | 0.6985 | 0.0006 | 0.0001 | 0.0    | 0.6497  | 6.3733  |
-| 1.193         | 2.4   | 6400  | 1.2418          | 68.252  | 6.8999 | 67.7704 | 68.023    | 80.4684  | 80.4684 | 80.4684 | 80.4684   | 0.6985 | 0.0006 | 0.0001 | 0.0    | 0.6506  | 6.3623  |
-| 1.1649        | 2.55  | 6800  | 1.2403          | 68.5799 | 6.7505 | 68.0777 | 68.3039   | 80.9368  | 80.9368 | 80.9368 | 80.9368   | 0.6993 | 0.0006 | 0.0001 | 0.0    | 0.6515  | 6.4743  |
-| 1.1488        | 2.7   | 7200  | 1.2401          | 68.6737 | 6.9022 | 68.2238 | 68.4258   | 81.0304  | 81.0304 | 81.0304 | 81.0304   | 0.7001 | 0.0006 | 0.0001 | 0.0    | 0.6523  | 6.4573  |
-| 1.1703        | 2.85  | 7600  | 1.2384          | 68.9471 | 6.7776 | 68.4667 | 68.6837   | 81.5457  | 81.5457 | 81.5457 | 81.5457   | 0.7004 | 0.0006 | 0.0001 | 0.0    | 0.6544  | 6.5267  |
-| 1.1763        | 3.0   | 8000  | 1.2348          | 68.7204 | 6.8157 | 68.2282 | 68.4466   | 81.1241  | 81.1241 | 81.1241 | 81.1241   | 0.7006 | 0.0006 | 0.0001 | 0.0    | 0.6527  | 6.4737  |
-| 1.0858        | 3.15  | 8400  | 1.2560          | 68.9519 | 6.9424 | 68.4982 | 68.6892   | 81.2646  | 81.2646 | 81.2646 | 81.2646   | 0.7025 | 0.0006 | 0.0001 | 0.0    | 0.6536  | 6.4767  |
-| 1.103         | 3.3   | 8800  | 1.2461          | 69.2609 | 6.8552 | 68.8244 | 69.0129   | 81.8267  | 81.8267 | 81.8267 | 81.8267   | 0.7064 | 0.0006 | 0.0001 | 0.0    | 0.6582  | 6.4533  |
-| 1.1183        | 3.45  | 9200  | 1.2507          | 68.8282 | 6.903  | 68.3923 | 68.5846   | 81.2178  | 81.2178 | 81.2178 | 81.2178   | 0.7018 | 0.0006 | 0.0001 | 0.0    | 0.6533  | 6.4797  |
-| 1.07          | 3.6   | 9600  | 1.2511          | 69.1742 | 6.8362 | 68.7377 | 68.906    | 81.733   | 81.733  | 81.733  | 81.733    | 0.7061 | 0.0006 | 0.0001 | 0.0    | 0.6576  | 6.4547  |
-| 1.0723        | 3.75  | 10000 | 1.2527          | 69.1098 | 6.7762 | 68.6426 | 68.8416   | 81.6862  | 81.6862 | 81.6862 | 81.6862   | 0.7052 | 0.0006 | 0.0001 | 0.0    | 0.6573  | 6.4493  |
-| 1.117         | 3.9   | 10400 | 1.2512          | 69.1469 | 6.7883 | 68.6792 | 68.9055   | 81.733   | 81.733  | 81.733  | 81.733    | 0.7051 | 0.0006 | 0.0001 | 0.0    | 0.6573  | 6.469   |
 ### Framework versions
-- Transformers 4.21.1
-- Pytorch 1.12.0
 - Datasets 2.3.2
 - Tokenizers 0.12.1

 tags:
 - generated_from_trainer
 model-index:
+- name: DIAL-FLANT5-XL
   results: []
+widget:
+ - text: "Instruction: Edit the provided response into a response that is fluent and coherent to the dialogue context. \n\nInput: [CONTEXT] How may I help you? [ENDOFTURN] I left a suitcase on the train to London the other day. [RESPONSE] Can describe itit , sir ? It will help us find [ENDOFDIALOGUE] [QUESTION] Given this context and response provided, the edited response is"
+ - text: "Instruction: Generate a response that starts with the provided initial phrase. \n\nInput: [INITIAL_PHRASE] Please describe [CONTEXT] How may I help you? [ENDOFTURN] I left a suitcase on the train to London the other day. [ENDOFDIALOGUE] [QUESTION] A response with the provided initial phrase is"
+ - text: "Instruction: Generate a response that starts with the provided initial phrase and contains the provided keywords. \n\nInput: [INITIAL PHRASE] Please describe [KEYWORDS] color, any documents [CONTEXT] How may I help you? [ENDOFTURN] I left a suitcase on the train to London the other day. [ENDOFDIALOGUE] [QUESTION] A response with the provided initial phrase and keywords is"
+ - text: "Instruction: What is the intent of the response \n\nInput: [CONTEXT] How may I help you? [RESPONSE] I left a suitcase on the train to London the other day. [ENDOFDIALOGUE] [OPTIONS] booking, reservation change, checkout, lost&found, time information, security, schedules [QUESTION] The intent of the response is"
+ - text: "Instruction: Generate a summary for the following dialog context. \n\nInput: [CONTEXT] Ann: Wanna go out? [ENDOFTURN] Kate: Not really, I feel sick. [ENDOFTURN] Ann: Drink mint tea, they say it helps. Ok, so we'll meet up another time. Take care! [ENDOFTURN] Kate: Thanks! [ENDOFDIALOGUE] [QUESTION] For this dialogue, the summary is: "
+ - text: "Instruction: Consider the context of the conversation and a document and generate an answer accordingly \n\nInput:  [CONTEXT] How may I help you? [ENDOFTURN] I left a suitcase on the train to London the other day. [ENDOFDIALOGUE] [QUESTION] What is the response of the following question: Where was the person going to?"
+ - text: "Instruction: Generate a response using the provided background knowledge. \n\nInput: [KNOWLEDGE] Emailid for cases related to lost and found is x@gmail.com [CONTEXT] How may I help you? [ENDOFTURN] I left a suitcase on the train to London the other day. [ENDOFDIALOGUE] [QUESTION] Generate a response using the information from the background knowledge."
 ---
+# InstructDial
+Instruction tuning is an emergent paradigm in NLP wherein natural language instructions are leveraged with language models to induce zero-shot performance on unseen tasks. Instructions have been shown to enable good performance on unseen tasks and datasets in both large and small language models. Dialogue is an especially interesting area to explore instruction tuning because dialogue systems perform multiple kinds of tasks related to language (e.g., natural language understanding and generation, domain-specific interaction), yet instruction tuning has not been systematically explored for dialogue-related tasks. We introduce InstructDial, an instruction tuning framework for dialogue, which consists of a repository of 48 diverse dialogue tasks in a unified text-to-text format created from 59 openly available dialogue datasets. Next, we explore cross-task generalization ability on models tuned on InstructDial across diverse dialogue tasks. Our analysis reveals that InstructDial enables good zero-shot performance on unseen datasets and tasks such as dialogue evaluation and intent detection, and even better performance in a few-shot setting. To ensure that models adhere to instructions, we introduce novel meta-tasks. We establish benchmark zero-shot and few-shot performance of models trained using the proposed framework on multiple dialogue tasks.
+[Paper](https://arxiv.org/abs/2205.12673)
+[GIT] (https://github.com/prakharguptaz/Instructdial)
+# DIAL-FLANT5-XL
+DIAL-FLANT5-XL is a 3B model trained on InstructDial tasks. This model is a fine-tuned version of google/flan-t5-xl model
 ## Model description
 ## Training and evaluation data
+All tasks in InstructDial framework (including all dialogue eval tasks)
 ## Training procedure
 The following hyperparameters were used during training:
 - learning_rate: 5e-05
+- train_batch_size: 9
+- eval_batch_size: 9
 - seed: 42
 - distributed_type: multi-GPU
+- num_devices: 8
+- total_train_batch_size: 72
+- total_eval_batch_size: 72
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
+- num_epochs: 3.0
 ### Framework versions
+- Transformers 4.20.1
+- Pytorch 1.11.0
 - Datasets 2.3.2
 - Tokenizers 0.12.1