wmgifford vijaye12 commited on
Commit
f494e7a
1 Parent(s): 350aa94

Update ReadMe (#6)

Browse files

- Update ReadMe (549da1efcb4066c6b9a705a2a78b73ac275a8015)
- Update README.md (96e1918907f5695f13a2c10b3ef04407e64d2c76)


Co-authored-by: Vijay Ekambaram <[email protected]>

Files changed (1) hide show
  1. README.md +12 -6
README.md CHANGED
@@ -16,8 +16,8 @@ forecasters, pre-trained on publicly available time series data with various aug
16
  fine-tuned for multi-variate forecasts with just 5% of the training data to be competitive. Refer to our [paper](https://arxiv.org/pdf/2401.03955v5.pdf) for more details.
17
 
18
 
19
- **The current open-source version supports point forecasting use-cases ranging from minutely to hourly resolutions
20
- (Ex. 10 min, 15 min, 1 hour, etc.)**
21
 
22
  **Note that zeroshot, fine-tuning and inference tasks using TTM can easily be executed in 1 GPU machine or in laptops too!!**
23
 
@@ -35,6 +35,12 @@ Stay tuned for the release of the model weights for these newer variants.
35
  - Script for Finetuning with cross-channel correlation support - to be added soon
36
 
37
 
 
 
 
 
 
 
38
  ## Benchmark Highlights:
39
 
40
  - TTM (with less than 1 Million parameters) outperforms the following popular Pre-trained SOTAs demanding several hundred Million to Billions of parameters [paper](https://arxiv.org/pdf/2401.03955v5.pdf):
@@ -102,10 +108,7 @@ time-series variates, a critical capability lacking in existing counterparts.
102
  In addition, TTM also supports exogenous infusion and categorical data which is not released as part of this version.
103
  Stay tuned for these extended features.
104
 
105
- ## Recommended Use
106
- 1. Users have to externally standard scale their data independently for every channel before feeding it to the model (Refer to [TSP](https://github.com/IBM/tsfm/blob/main/tsfm_public/toolkit/time_series_preprocessor.py), our data processing utility for data scaling.)
107
- 2. Enabling any upsampling or prepending zeros to virtually increase the context length for shorter-length datasets is not recommended and will
108
- impact the model performance.
109
 
110
 
111
  ### Model Sources
@@ -114,6 +117,9 @@ Stay tuned for these extended features.
114
  - **Paper:** https://arxiv.org/pdf/2401.03955v5.pdf
115
  - **Paper (Newer variants, extended benchmarks):** https://arxiv.org/pdf/2401.03955.pdf
116
 
 
 
 
117
 
118
  ## Uses
119
 
 
16
  fine-tuned for multi-variate forecasts with just 5% of the training data to be competitive. Refer to our [paper](https://arxiv.org/pdf/2401.03955v5.pdf) for more details.
17
 
18
 
19
+ **The current open-source version supports point forecasting use-cases specifically ranging from minutely to hourly resolutions
20
+ (Ex. 10 min, 15 min, 1 hour.).**
21
 
22
  **Note that zeroshot, fine-tuning and inference tasks using TTM can easily be executed in 1 GPU machine or in laptops too!!**
23
 
 
35
  - Script for Finetuning with cross-channel correlation support - to be added soon
36
 
37
 
38
+ ## Recommended Use
39
+ 1. Users have to externally standard scale their data independently for every channel before feeding it to the model (Refer to [TSP](https://github.com/IBM/tsfm/blob/main/tsfm_public/toolkit/time_series_preprocessor.py), our data processing utility for data scaling.)
40
+ 2. The current open-source version supports only minutely and hourly resolutions(Ex. 10 min, 15 min, 1 hour.). Other lower resolutions (say weekly, or monthly) are currently not supported in this version, as the model needs a minimum context length of 512 or 1024.
41
+ 3. Enabling any upsampling or prepending zeros to virtually increase the context length for shorter-length datasets is not recommended and will
42
+ impact the model performance.
43
+
44
  ## Benchmark Highlights:
45
 
46
  - TTM (with less than 1 Million parameters) outperforms the following popular Pre-trained SOTAs demanding several hundred Million to Billions of parameters [paper](https://arxiv.org/pdf/2401.03955v5.pdf):
 
108
  In addition, TTM also supports exogenous infusion and categorical data which is not released as part of this version.
109
  Stay tuned for these extended features.
110
 
111
+
 
 
 
112
 
113
 
114
  ### Model Sources
 
117
  - **Paper:** https://arxiv.org/pdf/2401.03955v5.pdf
118
  - **Paper (Newer variants, extended benchmarks):** https://arxiv.org/pdf/2401.03955.pdf
119
 
120
+ ### External Blogs on TTM
121
+ - https://aihorizonforecast.substack.com/p/tiny-time-mixersttms-powerful-zerofew
122
+ - https://medium.com/@david.proietti_17/predicting-venetian-lagoon-tide-levels-with-multivariate-time-series-modeling-8bafdf229588
123
 
124
  ## Uses
125